From be698daabb52a7f80faed6e5f6e357c41f9d96bc Mon Sep 17 00:00:00 2001 From: Mikhail Chusavitin Date: Tue, 14 Apr 2026 16:27:29 +0300 Subject: [PATCH] upgrade benchmark version and remove obsolete results; --- gpu-benchmark-20260410-184145/result.json | 579 ---------- gpu-benchmark-20260410-193819/result.json | 585 ---------- gpu-benchmark-20260411-085938/result.json | 577 ---------- gpu-benchmark-20260413-040542/result.json | 190 ---- gpu-benchmark-20260413-042023/result.json | 339 ------ gpu-benchmark-20260413-043509/result.json | 485 -------- gpu-benchmark-20260413-045001/result.json | 631 ----------- gpu-benchmark-20260413-050458/result.json | 776 ------------- gpu-benchmark-20260413-052002/result.json | 922 --------------- gpu-benchmark-20260413-053511/result.json | 1068 ------------------ gpu-benchmark-20260413-055026/result.json | 1234 --------------------- gpu-benchmark-20260414-152824/result.json | 141 +++ gpu-benchmark-20260414-152949/result.json | 385 +++++++ gpu-benchmark-20260414-154657/result.json | 735 ++++++++++++ 14 files changed, 1261 insertions(+), 7386 deletions(-) delete mode 100755 gpu-benchmark-20260410-184145/result.json delete mode 100755 gpu-benchmark-20260410-193819/result.json delete mode 100755 gpu-benchmark-20260411-085938/result.json delete mode 100644 gpu-benchmark-20260413-040542/result.json delete mode 100644 gpu-benchmark-20260413-042023/result.json delete mode 100644 gpu-benchmark-20260413-043509/result.json delete mode 100644 gpu-benchmark-20260413-045001/result.json delete mode 100644 gpu-benchmark-20260413-050458/result.json delete mode 100644 gpu-benchmark-20260413-052002/result.json delete mode 100644 gpu-benchmark-20260413-053511/result.json delete mode 100644 gpu-benchmark-20260413-055026/result.json create mode 100644 gpu-benchmark-20260414-152824/result.json create mode 100644 gpu-benchmark-20260414-152949/result.json create mode 100644 gpu-benchmark-20260414-154657/result.json diff --git a/gpu-benchmark-20260410-184145/result.json b/gpu-benchmark-20260410-184145/result.json deleted file mode 100755 index 720a48e..0000000 --- a/gpu-benchmark-20260410-184145/result.json +++ /dev/null @@ -1,579 +0,0 @@ -{ - "benchmark_version": "1", - "generated_at": "2026-04-10T18:41:45.128862028Z", - "hostname": "debian", - "server_model": "G5500 V7", - "benchmark_profile": "standard", - "overall_status": "PARTIAL", - "selected_gpu_indices": [ - 0, - 1, - 2, - 3 - ], - "findings": [ - "All 4 GPU(s) passed the benchmark.", - "Environment normalization was partial; compare results with caution.", - "GPU 0 ran without full benchmark normalization.", - "GPU 1 ran without full benchmark normalization.", - "GPU 2 showed unstable clocks/power over the benchmark window.", - "GPU 2 ran without full benchmark normalization.", - "GPU 3 ran without full benchmark normalization.", - "GPU power reporting may be unreliable: server delta 1310 W vs GPU-reported 2088 W (ratio 0.63). GPU telemetry likely over-reports actual consumption." - ], - "warnings": [ - "gpu inventory query failed: nvidia-smi gpu info (index,uuid,name,pci.bus_id,vbios_version): exit status 2" - ], - "normalization": { - "status": "partial", - "gpus": [ - { - "index": 0, - "persistence_mode": "applied", - "gpu_clock_lock_status": "skipped", - "notes": [ - "graphics clock lock skipped: gpu inventory unavailable or MaxGraphicsClockMHz=0" - ] - }, - { - "index": 1, - "persistence_mode": "applied", - "gpu_clock_lock_status": "skipped", - "notes": [ - "graphics clock lock skipped: gpu inventory unavailable or MaxGraphicsClockMHz=0" - ] - }, - { - "index": 2, - "persistence_mode": "applied", - "gpu_clock_lock_status": "skipped", - "notes": [ - "graphics clock lock skipped: gpu inventory unavailable or MaxGraphicsClockMHz=0" - ] - }, - { - "index": 3, - "persistence_mode": "applied", - "gpu_clock_lock_status": "skipped", - "notes": [ - "graphics clock lock skipped: gpu inventory unavailable or MaxGraphicsClockMHz=0" - ] - } - ] - }, - "gpus": [ - { - "index": 0, - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "baseline": { - "duration_sec": 15.64054741, - "samples": 16, - "avg_temp_c": 32, - "p95_temp_c": 32, - "avg_power_w": 35.71249999999999, - "p95_power_w": 35.9375, - "avg_graphics_clock_mhz": 180, - "p95_graphics_clock_mhz": 180, - "avg_memory_clock_mhz": 405, - "p95_memory_clock_mhz": 405, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 0.38374434928010487, - "temp_cv_pct": 0, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 517.039794776, - "samples": 517, - "avg_temp_c": 71.82978723404256, - "p95_temp_c": 73, - "avg_power_w": 526.566344294004, - "p95_power_w": 537.342, - "avg_graphics_clock_mhz": 2421.6382978723404, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 99.52417794970987, - "avg_mem_usage_pct": 39.276595744680854, - "clock_cv_pct": 0.33928654551992987, - "power_cv_pct": 6.134300387584527, - "temp_cv_pct": 1.6164905348919567, - "clock_drift_pct": 0.059851874611918786 - }, - "cooldown": { - "duration_sec": 120.692879388, - "samples": 117, - "avg_temp_c": 45.46153846153846, - "p95_temp_c": 54.19999999999999, - "avg_power_w": 40.62282051282055, - "p95_power_w": 55.827999999999996, - "avg_graphics_clock_mhz": 200.30769230769232, - "p95_graphics_clock_mhz": 180, - "avg_memory_clock_mhz": 1643.5641025641025, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 95.08186420156572, - "power_cv_pct": 13.397079812097019, - "temp_cv_pct": 8.815261908432287, - "clock_drift_pct": 31.35192004208311 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 0, - "thermal_sustain_score": 100, - "stability_score": 86.25454929352739, - "interconnect_score": 0, - "composite_score": 432.9908197154495 - }, - "degradation_reasons": [ - "normalization_partial" - ], - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition" - }, - { - "index": 1, - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "baseline": { - "duration_sec": 15.648620713, - "samples": 16, - "avg_temp_c": 33, - "p95_temp_c": 33, - "avg_power_w": 34.43125000000001, - "p95_power_w": 34.83, - "avg_graphics_clock_mhz": 180, - "p95_graphics_clock_mhz": 180, - "avg_memory_clock_mhz": 405, - "p95_memory_clock_mhz": 405, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 0.49361738151609885, - "temp_cv_pct": 0, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 517.039712514, - "samples": 517, - "avg_temp_c": 72.47969052224371, - "p95_temp_c": 74, - "avg_power_w": 525.4105029013539, - "p95_power_w": 533.974, - "avg_graphics_clock_mhz": 2422.015473887814, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 99.61315280464217, - "avg_mem_usage_pct": 39.15860735009671, - "clock_cv_pct": 0.014512672044927075, - "power_cv_pct": 5.368261116854865, - "temp_cv_pct": 1.8072124291831093, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.671014238, - "samples": 117, - "avg_temp_c": 45.38461538461539, - "p95_temp_c": 54, - "avg_power_w": 38.87683760683761, - "p95_power_w": 53.337999999999994, - "avg_graphics_clock_mhz": 189.5982905982906, - "p95_graphics_clock_mhz": 180, - "avg_memory_clock_mhz": 1643.5641025641025, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 46.5414056859924, - "power_cv_pct": 13.395397467444464, - "temp_cv_pct": 8.157534756058777, - "clock_drift_pct": 17.934183593134943 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 0, - "thermal_sustain_score": 100, - "stability_score": 89.20542707811056, - "interconnect_score": 0, - "composite_score": 436.2987496625886 - }, - "degradation_reasons": [ - "normalization_partial" - ], - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition" - }, - { - "index": 2, - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "baseline": { - "duration_sec": 15.646607951, - "samples": 16, - "avg_temp_c": 28.625, - "p95_temp_c": 29, - "avg_power_w": 35.14999999999999, - "p95_power_w": 35.15, - "avg_graphics_clock_mhz": 180, - "p95_graphics_clock_mhz": 180, - "avg_memory_clock_mhz": 405, - "p95_memory_clock_mhz": 405, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 2.0214587077101006e-14, - "temp_cv_pct": 1.6912591031473436, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 517.039507345, - "samples": 517, - "avg_temp_c": 70.29013539651838, - "p95_temp_c": 73, - "avg_power_w": 519.9781431334626, - "p95_power_w": 529.014, - "avg_graphics_clock_mhz": 2418.79497098646, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 99.42166344294004, - "avg_mem_usage_pct": 39.353965183752415, - "clock_cv_pct": 3.0099394503204846, - "power_cv_pct": 6.244661163316872, - "temp_cv_pct": 2.439779963276932, - "clock_drift_pct": 0.530345220491747 - }, - "cooldown": { - "duration_sec": 120.657452543, - "samples": 117, - "avg_temp_c": 44.623931623931625, - "p95_temp_c": 52.19999999999999, - "avg_power_w": 40.580256410256354, - "p95_power_w": 54.64799999999999, - "avg_graphics_clock_mhz": 186.07692307692307, - "p95_graphics_clock_mhz": 180, - "avg_memory_clock_mhz": 1643.5641025641025, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 29.666924389599842, - "power_cv_pct": 12.772094703863496, - "temp_cv_pct": 8.475122734640708, - "clock_drift_pct": 12.241118033338953 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 0, - "thermal_sustain_score": 100, - "stability_score": 74.41022943110082, - "interconnect_score": 0, - "composite_score": 419.7133534031841 - }, - "degradation_reasons": [ - "variance_too_high", - "normalization_partial" - ], - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition" - }, - { - "index": 3, - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "baseline": { - "duration_sec": 15.640129253, - "samples": 16, - "avg_temp_c": 30.375, - "p95_temp_c": 31, - "avg_power_w": 34.65437499999999, - "p95_power_w": 34.66, - "avg_graphics_clock_mhz": 179.5, - "p95_graphics_clock_mhz": 180, - "avg_memory_clock_mhz": 405, - "p95_memory_clock_mhz": 405, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 1.0788254446260215, - "power_cv_pct": 0.01431502864571969, - "temp_cv_pct": 1.5938203070812416, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 517.039084611, - "samples": 517, - "avg_temp_c": 70.26305609284333, - "p95_temp_c": 71, - "avg_power_w": 515.7172727272728, - "p95_power_w": 522.364, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 99.61315280464217, - "avg_mem_usage_pct": 39.22437137330754, - "clock_cv_pct": 0, - "power_cv_pct": 5.764080613784593, - "temp_cv_pct": 1.5613533584243726, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.710135838, - "samples": 117, - "avg_temp_c": 44.58974358974359, - "p95_temp_c": 52, - "avg_power_w": 39.0471794871795, - "p95_power_w": 52.519999999999996, - "avg_graphics_clock_mhz": 186.97435897435898, - "p95_graphics_clock_mhz": 180, - "avg_memory_clock_mhz": 1643.5641025641025, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 27.466449046011803, - "power_cv_pct": 12.490801537100193, - "temp_cv_pct": 8.28381511460339, - "clock_drift_pct": 13.747521480502309 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 0, - "thermal_sustain_score": 100, - "stability_score": 88.47183877243081, - "interconnect_score": 0, - "composite_score": 435.47639817859726 - }, - "degradation_reasons": [ - "normalization_partial" - ], - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition" - } - ], - "server_power": { - "available": true, - "idle_w": 632, - "loaded_w": 1942.2352941176468, - "delta_w": 1310.2352941176468, - "gpu_reported_sum_w": 2087.6722630560935, - "reporting_ratio": 0.6276058351226187 - } -} diff --git a/gpu-benchmark-20260410-193819/result.json b/gpu-benchmark-20260410-193819/result.json deleted file mode 100755 index 904861a..0000000 --- a/gpu-benchmark-20260410-193819/result.json +++ /dev/null @@ -1,585 +0,0 @@ -{ - "benchmark_version": "1", - "generated_at": "2026-04-10T19:38:19.724808089Z", - "hostname": "debian", - "server_model": "G5500 V7", - "benchmark_profile": "standard", - "parallel_gpus": true, - "overall_status": "PARTIAL", - "selected_gpu_indices": [ - 0, - 1, - 2, - 3 - ], - "findings": [ - "All 4 GPU(s) passed the benchmark.", - "Environment normalization was partial; compare results with caution.", - "GPU 0 showed unstable clocks/power over the benchmark window.", - "GPU 0 ran without full benchmark normalization.", - "GPU 1 showed unstable clocks/power over the benchmark window.", - "GPU 1 ran without full benchmark normalization.", - "GPU 2 showed unstable clocks/power over the benchmark window.", - "GPU 2 ran without full benchmark normalization.", - "GPU 3 showed unstable clocks/power over the benchmark window.", - "GPU 3 ran without full benchmark normalization." - ], - "warnings": [ - "gpu inventory query failed: nvidia-smi gpu info (index,uuid,name,pci.bus_id,vbios_version): exit status 2" - ], - "normalization": { - "status": "partial", - "gpus": [ - { - "index": 0, - "persistence_mode": "applied", - "gpu_clock_lock_status": "skipped", - "notes": [ - "graphics clock lock skipped: gpu inventory unavailable or MaxGraphicsClockMHz=0" - ] - }, - { - "index": 1, - "persistence_mode": "applied", - "gpu_clock_lock_status": "skipped", - "notes": [ - "graphics clock lock skipped: gpu inventory unavailable or MaxGraphicsClockMHz=0" - ] - }, - { - "index": 2, - "persistence_mode": "applied", - "gpu_clock_lock_status": "skipped", - "notes": [ - "graphics clock lock skipped: gpu inventory unavailable or MaxGraphicsClockMHz=0" - ] - }, - { - "index": 3, - "persistence_mode": "applied", - "gpu_clock_lock_status": "skipped", - "notes": [ - "graphics clock lock skipped: gpu inventory unavailable or MaxGraphicsClockMHz=0" - ] - } - ] - }, - "gpus": [ - { - "index": 0, - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "baseline": { - "duration_sec": 15.708738197, - "samples": 16, - "avg_temp_c": 31.5, - "p95_temp_c": 32, - "avg_power_w": 35.63999999999999, - "p95_power_w": 35.64, - "avg_graphics_clock_mhz": 180, - "p95_graphics_clock_mhz": 180, - "avg_memory_clock_mhz": 405, - "p95_memory_clock_mhz": 405, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 1.9936664864200346e-14, - "temp_cv_pct": 1.5873015873015872, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 521.038266969, - "samples": 521, - "avg_temp_c": 71.51247600767755, - "p95_temp_c": 72, - "avg_power_w": 521.897197696737, - "p95_power_w": 532, - "avg_graphics_clock_mhz": 2405.7063339731285, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 98.69481765834932, - "avg_mem_usage_pct": 38.351247600767756, - "clock_cv_pct": 7.7140057202952, - "power_cv_pct": 9.892799596716745, - "temp_cv_pct": 2.580458060863098, - "clock_drift_pct": 2.6961189099917497 - }, - "cooldown": { - "duration_sec": 120.171037413, - "samples": 116, - "avg_temp_c": 44.310344827586206, - "p95_temp_c": 51, - "avg_power_w": 39.81387931034485, - "p95_power_w": 54.0825, - "avg_graphics_clock_mhz": 179.6551724137931, - "p95_graphics_clock_mhz": 180, - "avg_memory_clock_mhz": 1237.8275862068965, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0.9043546249051633, - "power_cv_pct": 10.379008691678044, - "temp_cv_pct": 6.727580324493612, - "clock_drift_pct": 0.1537279016141304 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 0, - "thermal_sustain_score": 100, - "stability_score": 43.96614010540221, - "interconnect_score": 0, - "composite_score": 385.5855710463551 - }, - "degradation_reasons": [ - "variance_too_high", - "normalization_partial" - ], - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition" - }, - { - "index": 1, - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "baseline": { - "duration_sec": 15.708738197, - "samples": 16, - "avg_temp_c": 34, - "p95_temp_c": 34, - "avg_power_w": 34.82999999999999, - "p95_power_w": 34.83, - "avg_graphics_clock_mhz": 179.5, - "p95_graphics_clock_mhz": 180, - "avg_memory_clock_mhz": 405, - "p95_memory_clock_mhz": 405, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 1.0788254446260215, - "power_cv_pct": 2.040030823313524e-14, - "temp_cv_pct": 0, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 521.038266969, - "samples": 521, - "avg_temp_c": 71.71593090211132, - "p95_temp_c": 72, - "avg_power_w": 518.5774088291746, - "p95_power_w": 528.2, - "avg_graphics_clock_mhz": 2410.6295585412668, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 98.85220729366603, - "avg_mem_usage_pct": 38.35892514395393, - "clock_cv_pct": 6.24737613351838, - "power_cv_pct": 9.166896001630583, - "temp_cv_pct": 2.232861903238498, - "clock_drift_pct": 1.8814711300260394 - }, - "cooldown": { - "duration_sec": 120.171037413, - "samples": 116, - "avg_temp_c": 45.741379310344826, - "p95_temp_c": 52.25, - "avg_power_w": 38.606724137931074, - "p95_power_w": 52.75, - "avg_graphics_clock_mhz": 179.86206896551724, - "p95_graphics_clock_mhz": 180, - "avg_memory_clock_mhz": 1446.0344827586207, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0.5789750333796595, - "power_cv_pct": 11.580300308869425, - "temp_cv_pct": 6.546796049282795, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 0, - "thermal_sustain_score": 100, - "stability_score": 52.91376120261324, - "interconnect_score": 0, - "composite_score": 395.6158420178446 - }, - "degradation_reasons": [ - "variance_too_high", - "normalization_partial" - ], - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition" - }, - { - "index": 2, - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "baseline": { - "duration_sec": 15.708738197, - "samples": 16, - "avg_temp_c": 30, - "p95_temp_c": 30, - "avg_power_w": 35.59374999999999, - "p95_power_w": 35.64, - "avg_graphics_clock_mhz": 180, - "p95_graphics_clock_mhz": 180, - "avg_memory_clock_mhz": 405, - "p95_memory_clock_mhz": 405, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 0.3638624937385677, - "temp_cv_pct": 0, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 521.038266969, - "samples": 521, - "avg_temp_c": 67.8042226487524, - "p95_temp_c": 69, - "avg_power_w": 511.71065259117063, - "p95_power_w": 521.61, - "avg_graphics_clock_mhz": 2406.5124760076774, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 98.6852207293666, - "avg_mem_usage_pct": 38.618042226487525, - "clock_cv_pct": 7.369275290697701, - "power_cv_pct": 9.95377772121058, - "temp_cv_pct": 2.709178533043663, - "clock_drift_pct": 2.5627262910499864 - }, - "cooldown": { - "duration_sec": 120.171037413, - "samples": 116, - "avg_temp_c": 41.293103448275865, - "p95_temp_c": 47.25, - "avg_power_w": 39.30818965517238, - "p95_power_w": 52.955, - "avg_graphics_clock_mhz": 179.86206896551724, - "p95_graphics_clock_mhz": 180, - "avg_memory_clock_mhz": 1237.8275862068965, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0.5789750333796595, - "power_cv_pct": 9.84500039444727, - "temp_cv_pct": 6.8127582501704795, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 0, - "thermal_sustain_score": 100, - "stability_score": 45.48989081268807, - "interconnect_score": 0, - "composite_score": 387.2936934982367 - }, - "degradation_reasons": [ - "variance_too_high", - "normalization_partial" - ], - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition" - }, - { - "index": 3, - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "baseline": { - "duration_sec": 15.708738197, - "samples": 16, - "avg_temp_c": 34, - "p95_temp_c": 34, - "avg_power_w": 35.513124999999995, - "p95_power_w": 35.64, - "avg_graphics_clock_mhz": 180, - "p95_graphics_clock_mhz": 180, - "avg_memory_clock_mhz": 405, - "p95_memory_clock_mhz": 405, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 0.4941532782477418, - "temp_cv_pct": 0, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 521.038266969, - "samples": 521, - "avg_temp_c": 68.54126679462571, - "p95_temp_c": 69, - "avg_power_w": 509.9767370441462, - "p95_power_w": 519.72, - "avg_graphics_clock_mhz": 2404.0940499040307, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 98.7447216890595, - "avg_mem_usage_pct": 38.42610364683301, - "clock_cv_pct": 8.026350243134432, - "power_cv_pct": 9.808646251646744, - "temp_cv_pct": 2.580108116733557, - "clock_drift_pct": 1.5637794268167964 - }, - "cooldown": { - "duration_sec": 120.171037413, - "samples": 116, - "avg_temp_c": 42.077586206896555, - "p95_temp_c": 48.25, - "avg_power_w": 38.305517241379285, - "p95_power_w": 51.4, - "avg_graphics_clock_mhz": 179.79310344827587, - "p95_graphics_clock_mhz": 180, - "avg_memory_clock_mhz": 1341.9310344827586, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0.7062505807258133, - "power_cv_pct": 10.343052937274862, - "temp_cv_pct": 7.185134807174834, - "clock_drift_pct": 0.15349194167306543 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 0, - "thermal_sustain_score": 100, - "stability_score": 45.149747670535184, - "interconnect_score": 0, - "composite_score": 386.91239350264897 - }, - "degradation_reasons": [ - "variance_too_high", - "normalization_partial" - ], - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition" - } - ], - "server_power": { - "available": true, - "idle_w": 670.6666666666666, - "loaded_w": 3577.3980582524273, - "delta_w": 2906.731391585761, - "gpu_reported_sum_w": 2062.161996161228, - "reporting_ratio": 1.4095553099110167 - } -} diff --git a/gpu-benchmark-20260411-085938/result.json b/gpu-benchmark-20260411-085938/result.json deleted file mode 100755 index 751f5ef..0000000 --- a/gpu-benchmark-20260411-085938/result.json +++ /dev/null @@ -1,577 +0,0 @@ -{ - "benchmark_version": "1", - "generated_at": "2026-04-11T08:59:38.567052641Z", - "hostname": "debian", - "server_model": "G5500 V7", - "benchmark_profile": "stability", - "overall_status": "PARTIAL", - "selected_gpu_indices": [ - 0, - 1, - 2, - 3 - ], - "findings": [ - "All 4 GPU(s) passed the benchmark.", - "Environment normalization was partial; compare results with caution.", - "GPU 0 ran without full benchmark normalization.", - "GPU 1 ran without full benchmark normalization.", - "GPU 2 ran without full benchmark normalization.", - "GPU 3 ran without full benchmark normalization.", - "GPU power reporting may be unreliable: server delta 990 W vs GPU-reported 2084 W (ratio 0.47). GPU telemetry likely over-reports actual consumption." - ], - "warnings": [ - "gpu inventory query failed: nvidia-smi gpu info (index,uuid,name,pci.bus_id,vbios_version): exit status 2" - ], - "normalization": { - "status": "partial", - "gpus": [ - { - "index": 0, - "persistence_mode": "applied", - "gpu_clock_lock_status": "skipped", - "notes": [ - "graphics clock lock skipped: gpu inventory unavailable or MaxGraphicsClockMHz=0" - ] - }, - { - "index": 1, - "persistence_mode": "applied", - "gpu_clock_lock_status": "skipped", - "notes": [ - "graphics clock lock skipped: gpu inventory unavailable or MaxGraphicsClockMHz=0" - ] - }, - { - "index": 2, - "persistence_mode": "applied", - "gpu_clock_lock_status": "skipped", - "notes": [ - "graphics clock lock skipped: gpu inventory unavailable or MaxGraphicsClockMHz=0" - ] - }, - { - "index": 3, - "persistence_mode": "applied", - "gpu_clock_lock_status": "skipped", - "notes": [ - "graphics clock lock skipped: gpu inventory unavailable or MaxGraphicsClockMHz=0" - ] - } - ] - }, - "gpus": [ - { - "index": 0, - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "baseline": { - "duration_sec": 30.999116342, - "samples": 31, - "avg_temp_c": 50, - "p95_temp_c": 54.5, - "avg_power_w": 104.47096774193548, - "p95_power_w": 108.61, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 100, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 2.028975279257823, - "temp_cv_pct": 5.003224766539689, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 3652.033401705, - "samples": 3652, - "avg_temp_c": 70.00602409638554, - "p95_temp_c": 70, - "avg_power_w": 524.3030010952905, - "p95_power_w": 528.23, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 99.93729463307777, - "avg_mem_usage_pct": 39.491511500547645, - "clock_cv_pct": 0, - "power_cv_pct": 2.148355726733903, - "temp_cv_pct": 0.5358272117706183, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 300.134211129, - "samples": 291, - "avg_temp_c": 38.807560137457045, - "p95_temp_c": 46, - "avg_power_w": 38.135498281786944, - "p95_power_w": 39.61, - "avg_graphics_clock_mhz": 187.59106529209623, - "p95_graphics_clock_mhz": 180, - "avg_memory_clock_mhz": 944.4776632302405, - "p95_memory_clock_mhz": 405, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 64.75534616761331, - "power_cv_pct": 9.705769321538293, - "temp_cv_pct": 11.481450187501306, - "clock_drift_pct": 14.989814023789192 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 197664, - "teraops_per_sec": 185.23959771267073 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 197664, - "teraops_per_sec": 63.24483347797334 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 197664, - "teraops_per_sec": 282.12158593848665 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 530.6060171291307, - "power_sustain_score": 0, - "thermal_sustain_score": 100, - "stability_score": 95.70328854653219, - "interconnect_score": 0, - "composite_score": 419.9250918011493 - }, - "degradation_reasons": [ - "normalization_partial" - ], - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition" - }, - { - "index": 1, - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "baseline": { - "duration_sec": 30.024773184, - "samples": 30, - "avg_temp_c": 49, - "p95_temp_c": 49, - "avg_power_w": 99.5443333333333, - "p95_power_w": 99.59, - "avg_graphics_clock_mhz": 2423.3333333333335, - "p95_graphics_clock_mhz": 2430, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 100, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0.12302987496560051, - "power_cv_pct": 0.08327227979297677, - "temp_cv_pct": 0, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 3608.034869542, - "samples": 3608, - "avg_temp_c": 69.85864745011087, - "p95_temp_c": 70, - "avg_power_w": 519.3211474501111, - "p95_power_w": 523.5, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 99.94456762749445, - "avg_mem_usage_pct": 38.920731707317074, - "clock_cv_pct": 0, - "power_cv_pct": 1.9854404455976928, - "temp_cv_pct": 0.6438449280694504, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 300.694502766, - "samples": 291, - "avg_temp_c": 31.02405498281787, - "p95_temp_c": 49.5, - "avg_power_w": 34.87268041237119, - "p95_power_w": 37.97, - "avg_graphics_clock_mhz": 182.0996563573883, - "p95_graphics_clock_mhz": 180, - "avg_memory_clock_mhz": 902.979381443299, - "p95_memory_clock_mhz": 405, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 17.117532557949268, - "power_cv_pct": 11.261564470591717, - "temp_cv_pct": 22.686094189937133, - "clock_drift_pct": 4.900448166923811 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 194880, - "teraops_per_sec": 182.6305892941824 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 194880, - "teraops_per_sec": 62.35406117546666 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 194880, - "teraops_per_sec": 278.1480424745643 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 523.1326929442134, - "power_sustain_score": 0, - "thermal_sustain_score": 100, - "stability_score": 96.02911910880461, - "interconnect_score": 0, - "composite_score": 414.35155912742727 - }, - "degradation_reasons": [ - "normalization_partial" - ], - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition" - }, - { - "index": 2, - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "baseline": { - "duration_sec": 30.083934487, - "samples": 30, - "avg_temp_c": 40.06666666666667, - "p95_temp_c": 41, - "avg_power_w": 104.565, - "p95_power_w": 105.1175, - "avg_graphics_clock_mhz": 2317.5333333333333, - "p95_graphics_clock_mhz": 2321.4, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 100, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0.08610666252680013, - "power_cv_pct": 0.309119766163425, - "temp_cv_pct": 1.6968450960375328, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 3653.036306358, - "samples": 3653, - "avg_temp_c": 69.79742677251573, - "p95_temp_c": 71, - "avg_power_w": 521.0794552422657, - "p95_power_w": 525.36, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 99.9452504790583, - "avg_mem_usage_pct": 39.54913769504517, - "clock_cv_pct": 0, - "power_cv_pct": 1.9968664278288677, - "temp_cv_pct": 0.9113538767500047, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 300.206414655, - "samples": 290, - "avg_temp_c": 38.644827586206894, - "p95_temp_c": 48, - "avg_power_w": 38.240275862069, - "p95_power_w": 39.6455, - "avg_graphics_clock_mhz": 182.5448275862069, - "p95_graphics_clock_mhz": 180, - "avg_memory_clock_mhz": 946.3379310344827, - "p95_memory_clock_mhz": 405, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 17.07777076535198, - "power_cv_pct": 9.812483217019517, - "temp_cv_pct": 12.52523069358715, - "clock_drift_pct": 5.779589459892267 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 197648, - "teraops_per_sec": 185.22460341141502 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 197648, - "teraops_per_sec": 63.239714096924445 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 197648, - "teraops_per_sec": 282.09874948179737 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 530.5630669901368, - "power_sustain_score": 0, - "thermal_sustain_score": 100, - "stability_score": 96.00626714434226, - "interconnect_score": 0, - "composite_score": 420.21259928683537 - }, - "degradation_reasons": [ - "normalization_partial" - ], - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition" - }, - { - "index": 3, - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "baseline": { - "duration_sec": 30.126177597, - "samples": 30, - "avg_temp_c": 46.7, - "p95_temp_c": 47, - "avg_power_w": 98.56633333333333, - "p95_power_w": 98.83, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 100, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 0.18649396425550566, - "temp_cv_pct": 0.9812795920676312, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 3603.039321222, - "samples": 3603, - "avg_temp_c": 69.96364140993616, - "p95_temp_c": 71, - "avg_power_w": 518.9227643630296, - "p95_power_w": 523.27, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 99.93422148209825, - "avg_mem_usage_pct": 39.586455731335, - "clock_cv_pct": 0, - "power_cv_pct": 2.281713247433057, - "temp_cv_pct": 1.1128723340072593, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 300.670868947, - "samples": 290, - "avg_temp_c": 38.148275862068964, - "p95_temp_c": 47, - "avg_power_w": 38.17251724137949, - "p95_power_w": 39.61, - "avg_graphics_clock_mhz": 189.0103448275862, - "p95_graphics_clock_mhz": 180, - "avg_memory_clock_mhz": 946.3379310344827, - "p95_memory_clock_mhz": 405, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 70.2162030341821, - "power_cv_pct": 10.027005094761156, - "temp_cv_pct": 12.51736757591107, - "clock_drift_pct": 17.23078891803699 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 194896, - "teraops_per_sec": 182.6455835954381 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 194896, - "teraops_per_sec": 62.359180556515554 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 194896, - "teraops_per_sec": 278.17087893125347 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 523.1756430832071, - "power_sustain_score": 0, - "thermal_sustain_score": 100, - "stability_score": 95.43657350513388, - "interconnect_score": 0, - "composite_score": 413.76556728433667 - }, - "degradation_reasons": [ - "normalization_partial" - ], - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition" - } - ], - "server_power": { - "available": true, - "idle_w": 936, - "loaded_w": 1925.6498918250659, - "delta_w": 989.6498918250659, - "gpu_reported_sum_w": 2083.626368150697, - "reporting_ratio": 0.47496514104081927 - } -} diff --git a/gpu-benchmark-20260413-040542/result.json b/gpu-benchmark-20260413-040542/result.json deleted file mode 100644 index 1833570..0000000 --- a/gpu-benchmark-20260413-040542/result.json +++ /dev/null @@ -1,190 +0,0 @@ -{ - "benchmark_version": "1", - "generated_at": "2026-04-13T04:05:42.675718581Z", - "hostname": "debian", - "server_model": "G5500 V7", - "benchmark_profile": "standard", - "parallel_gpus": true, - "ramp_step": 1, - "ramp_total": 8, - "ramp_run_id": "ramp-20260413-035309", - "overall_status": "OK", - "selected_gpu_indices": [ - 0 - ], - "findings": [ - "All 1 GPU(s) passed the benchmark.", - "GPU 0 held clocks without observable throttle counters during steady state.", - "Server power delta 1188 W exceeds GPU-reported sum 514 W by 131%. Other components (CPU, NVMe, networking) may be drawing substantial power under GPU load." - ], - "normalization": { - "status": "full", - "gpus": [ - { - "index": 0, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - } - ] - }, - "host_config": { - "cpu_model": "INTEL(R) XEON(R) GOLD 6530", - "cpu_sockets": 2, - "cpu_cores": 64, - "cpu_threads": 128, - "mem_total_gib": 125.36544799804688 - }, - "cpu_load": { - "avg_pct": 1.1, - "max_pct": 4.5, - "p95_pct": 4, - "samples": 71, - "status": "ok" - }, - "gpus": [ - { - "index": 0, - "uuid": "GPU-8281bc14-c076-e306-75b1-e91596be45e7", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:4B:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.28, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 15.070173316, - "samples": 15, - "avg_temp_c": 57, - "p95_temp_c": 62.599999999999994, - "avg_power_w": 119.25133333333333, - "p95_power_w": 207.5479999999996, - "avg_graphics_clock_mhz": 2418, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0.6189673096400233, - "power_cv_pct": 74.77759859420686, - "temp_cv_pct": 6.077371254627639, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 518.070872587, - "samples": 518, - "avg_temp_c": 71.9073359073359, - "p95_temp_c": 72, - "avg_power_w": 514.3837065637066, - "p95_power_w": 520.4545, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 99.54826254826254, - "avg_mem_usage_pct": 38.46138996138996, - "clock_cv_pct": 0, - "power_cv_pct": 5.442205008113535, - "temp_cv_pct": 0.9844059147322263, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.076189386, - "samples": 113, - "avg_temp_c": 49.92920353982301, - "p95_temp_c": 56, - "avg_power_w": 89.37265486725663, - "p95_power_w": 94.67, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 2.53783367168321, - "temp_cv_pct": 4.882111393317778, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.95333333333333, - "thermal_sustain_score": 100, - "stability_score": 89.11558998377294, - "interconnect_score": 0, - "composite_score": 545.2083180519243 - } - } - ], - "server_power": { - "available": true, - "idle_w": 993.3333333333334, - "loaded_w": 2181.4117647058824, - "delta_w": 1188.078431372549, - "gpu_reported_sum_w": 514.3837065637066, - "reporting_ratio": 2.3097124115952243 - } -} \ No newline at end of file diff --git a/gpu-benchmark-20260413-042023/result.json b/gpu-benchmark-20260413-042023/result.json deleted file mode 100644 index 84455cd..0000000 --- a/gpu-benchmark-20260413-042023/result.json +++ /dev/null @@ -1,339 +0,0 @@ -{ - "benchmark_version": "1", - "generated_at": "2026-04-13T04:20:23.254080304Z", - "hostname": "debian", - "server_model": "G5500 V7", - "benchmark_profile": "standard", - "parallel_gpus": true, - "ramp_step": 2, - "ramp_total": 8, - "ramp_run_id": "ramp-20260413-035309", - "overall_status": "OK", - "selected_gpu_indices": [ - 0, - 1 - ], - "findings": [ - "All 2 GPU(s) passed the benchmark.", - "GPU 0 showed unstable clocks/power over the benchmark window.", - "GPU 1 showed unstable clocks/power over the benchmark window.", - "Server power delta 1672 W exceeds GPU-reported sum 1036 W by 61%. Other components (CPU, NVMe, networking) may be drawing substantial power under GPU load." - ], - "normalization": { - "status": "full", - "gpus": [ - { - "index": 0, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 1, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - } - ] - }, - "host_config": { - "cpu_model": "INTEL(R) XEON(R) GOLD 6530", - "cpu_sockets": 2, - "cpu_cores": 64, - "cpu_threads": 128, - "mem_total_gib": 125.36544799804688 - }, - "cpu_load": { - "avg_pct": 1.5, - "max_pct": 1.6, - "p95_pct": 1.6, - "samples": 71, - "status": "ok" - }, - "gpus": [ - { - "index": 0, - "uuid": "GPU-8281bc14-c076-e306-75b1-e91596be45e7", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:4B:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.22, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 15.057418789, - "samples": 15, - "avg_temp_c": 59.53333333333333, - "p95_temp_c": 65.6, - "avg_power_w": 118.17133333333335, - "p95_power_w": 195.23499999999967, - "avg_graphics_clock_mhz": 2419.5333333333333, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0.3814546146932904, - "power_cv_pct": 64.74043823099278, - "temp_cv_pct": 6.098655679368466, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 520.069532944, - "samples": 520, - "avg_temp_c": 71.74807692307692, - "p95_temp_c": 72, - "avg_power_w": 511.9014230769234, - "p95_power_w": 520.072, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 99.03846153846153, - "avg_mem_usage_pct": 38.41538461538462, - "clock_cv_pct": 0, - "power_cv_pct": 7.630331977379506, - "temp_cv_pct": 1.6849782906634136, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 121.009027315, - "samples": 114, - "avg_temp_c": 49.79824561403509, - "p95_temp_c": 54.349999999999994, - "avg_power_w": 89.54026315789477, - "p95_power_w": 93.49049999999998, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 1.7838271138649684, - "temp_cv_pct": 3.883484530370241, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.96333333333332, - "thermal_sustain_score": 100, - "stability_score": 84.73933604524099, - "interconnect_score": 0, - "composite_score": 539.0845072169755 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 1, - "uuid": "GPU-a66d6d68-9870-dfe0-1823-b38a141c21ae", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:4C:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.225, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 15.057418789, - "samples": 15, - "avg_temp_c": 56.333333333333336, - "p95_temp_c": 61.599999999999994, - "avg_power_w": 113.68933333333335, - "p95_power_w": 194.66699999999966, - "avg_graphics_clock_mhz": 2419.5333333333333, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0.3814546146932904, - "power_cv_pct": 71.92406919974427, - "temp_cv_pct": 5.675540264275408, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 520.069532944, - "samples": 520, - "avg_temp_c": 74.13461538461539, - "p95_temp_c": 75, - "avg_power_w": 523.9403653846153, - "p95_power_w": 532.881, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 98.86923076923077, - "avg_mem_usage_pct": 39.32884615384615, - "clock_cv_pct": 0, - "power_cv_pct": 8.668121242011786, - "temp_cv_pct": 1.9390814166506207, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 121.009027315, - "samples": 114, - "avg_temp_c": 51.63157894736842, - "p95_temp_c": 56, - "avg_power_w": 88.05026315789462, - "p95_power_w": 92.084, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 1.8363510924174664, - "temp_cv_pct": 3.5292319871148523, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.96249999999999, - "thermal_sustain_score": 100, - "stability_score": 82.66375751597643, - "interconnect_score": 0, - "composite_score": 536.175405738999 - }, - "degradation_reasons": [ - "variance_too_high" - ] - } - ], - "server_power": { - "available": true, - "idle_w": 1028, - "loaded_w": 2699.883495145631, - "delta_w": 1671.883495145631, - "gpu_reported_sum_w": 1035.8417884615387, - "reporting_ratio": 1.6140336427522965 - } -} \ No newline at end of file diff --git a/gpu-benchmark-20260413-043509/result.json b/gpu-benchmark-20260413-043509/result.json deleted file mode 100644 index e64eee2..0000000 --- a/gpu-benchmark-20260413-043509/result.json +++ /dev/null @@ -1,485 +0,0 @@ -{ - "benchmark_version": "1", - "generated_at": "2026-04-13T04:35:09.488332038Z", - "hostname": "debian", - "server_model": "G5500 V7", - "benchmark_profile": "standard", - "parallel_gpus": true, - "ramp_step": 3, - "ramp_total": 8, - "ramp_run_id": "ramp-20260413-035309", - "overall_status": "OK", - "selected_gpu_indices": [ - 0, - 1, - 2 - ], - "findings": [ - "All 3 GPU(s) passed the benchmark.", - "GPU 0 showed unstable clocks/power over the benchmark window.", - "GPU 1 showed unstable clocks/power over the benchmark window.", - "GPU 2 showed unstable clocks/power over the benchmark window.", - "Server power delta 2188 W exceeds GPU-reported sum 1565 W by 40%. Other components (CPU, NVMe, networking) may be drawing substantial power under GPU load." - ], - "normalization": { - "status": "full", - "gpus": [ - { - "index": 0, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 1, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 2, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - } - ] - }, - "host_config": { - "cpu_model": "INTEL(R) XEON(R) GOLD 6530", - "cpu_sockets": 2, - "cpu_cores": 64, - "cpu_threads": 128, - "mem_total_gib": 125.36544799804688 - }, - "cpu_load": { - "avg_pct": 1.9, - "max_pct": 2.5, - "p95_pct": 2.5, - "samples": 82, - "status": "ok" - }, - "gpus": [ - { - "index": 0, - "uuid": "GPU-8281bc14-c076-e306-75b1-e91596be45e7", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:4B:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.255, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 15.033289155, - "samples": 15, - "avg_temp_c": 59.2, - "p95_temp_c": 65.3, - "avg_power_w": 118.2433333333333, - "p95_power_w": 195.2919999999997, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 64.53847869928777, - "temp_cv_pct": 5.890403977757667, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 522.069202793, - "samples": 522, - "avg_temp_c": 71.73946360153256, - "p95_temp_c": 72, - "avg_power_w": 511.1453831417621, - "p95_power_w": 520.4285, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 98.83141762452107, - "avg_mem_usage_pct": 38.42145593869732, - "clock_cv_pct": 0, - "power_cv_pct": 8.76270864901095, - "temp_cv_pct": 2.1245135467552823, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.942911251, - "samples": 114, - "avg_temp_c": 49.43859649122807, - "p95_temp_c": 54, - "avg_power_w": 89.03017543859652, - "p95_power_w": 92.58, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 1.7468159750921186, - "temp_cv_pct": 3.6835591717577603, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.9575, - "thermal_sustain_score": 100, - "stability_score": 82.47458270197811, - "interconnect_score": 0, - "composite_score": 535.9061211105268 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 1, - "uuid": "GPU-a66d6d68-9870-dfe0-1823-b38a141c21ae", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:4C:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.165, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 15.033289155, - "samples": 15, - "avg_temp_c": 61, - "p95_temp_c": 66.6, - "avg_power_w": 112.94333333333337, - "p95_power_w": 177.84699999999975, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 57.05117016518504, - "temp_cv_pct": 5.678855106783204, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 522.069202793, - "samples": 522, - "avg_temp_c": 73.69731800766283, - "p95_temp_c": 74, - "avg_power_w": 521.9090804597699, - "p95_power_w": 532.379, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 98.46934865900383, - "avg_mem_usage_pct": 38.99233716475096, - "clock_cv_pct": 0, - "power_cv_pct": 9.915628616003222, - "temp_cv_pct": 2.3155813129932827, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.942911251, - "samples": 114, - "avg_temp_c": 50.75438596491228, - "p95_temp_c": 55, - "avg_power_w": 87.372894736842, - "p95_power_w": 90.94749999999999, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 1.7330150264667308, - "temp_cv_pct": 3.4872476446986864, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.97250000000001, - "thermal_sustain_score": 100, - "stability_score": 80.16874276799356, - "interconnect_score": 0, - "composite_score": 532.6876780928648 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 2, - "uuid": "GPU-82b32f5b-5fca-9674-a845-cfd5da365d09", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:4E:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.23, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 15.033289155, - "samples": 15, - "avg_temp_c": 53.733333333333334, - "p95_temp_c": 59.3, - "avg_power_w": 114.68533333333335, - "p95_power_w": 191.34599999999966, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 67.3613557450656, - "temp_cv_pct": 5.864169874985842, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 522.069202793, - "samples": 522, - "avg_temp_c": 73.5823754789272, - "p95_temp_c": 74, - "avg_power_w": 531.7479693486594, - "p95_power_w": 542.509, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 98.47892720306514, - "avg_mem_usage_pct": 38.74521072796935, - "clock_cv_pct": 0, - "power_cv_pct": 9.887326796340815, - "temp_cv_pct": 2.4110385580810907, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.942911251, - "samples": 114, - "avg_temp_c": 49.57017543859649, - "p95_temp_c": 54.349999999999994, - "avg_power_w": 90.77263157894727, - "p95_power_w": 94.967, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 2.017263598743592, - "temp_cv_pct": 4.05207099778575, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.96166666666666, - "thermal_sustain_score": 100, - "stability_score": 80.22534640731837, - "interconnect_score": 0, - "composite_score": 532.7578857315244 - }, - "degradation_reasons": [ - "variance_too_high" - ] - } - ], - "server_power": { - "available": true, - "idle_w": 1036, - "loaded_w": 3223.5728155339807, - "delta_w": 2187.5728155339807, - "gpu_reported_sum_w": 1564.8024329501914, - "reporting_ratio": 1.397986588894582 - } -} \ No newline at end of file diff --git a/gpu-benchmark-20260413-045001/result.json b/gpu-benchmark-20260413-045001/result.json deleted file mode 100644 index 3b8e5c3..0000000 --- a/gpu-benchmark-20260413-045001/result.json +++ /dev/null @@ -1,631 +0,0 @@ -{ - "benchmark_version": "1", - "generated_at": "2026-04-13T04:50:01.238832218Z", - "hostname": "debian", - "server_model": "G5500 V7", - "benchmark_profile": "standard", - "parallel_gpus": true, - "ramp_step": 4, - "ramp_total": 8, - "ramp_run_id": "ramp-20260413-035309", - "overall_status": "OK", - "selected_gpu_indices": [ - 0, - 1, - 2, - 3 - ], - "findings": [ - "All 4 GPU(s) passed the benchmark.", - "GPU 0 showed unstable clocks/power over the benchmark window.", - "GPU 1 showed unstable clocks/power over the benchmark window.", - "GPU 2 showed unstable clocks/power over the benchmark window.", - "GPU 3 showed unstable clocks/power over the benchmark window.", - "Server power delta 2690 W exceeds GPU-reported sum 2085 W by 29%. Other components (CPU, NVMe, networking) may be drawing substantial power under GPU load." - ], - "normalization": { - "status": "full", - "gpus": [ - { - "index": 0, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 1, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 2, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 3, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - } - ] - }, - "host_config": { - "cpu_model": "INTEL(R) XEON(R) GOLD 6530", - "cpu_sockets": 2, - "cpu_cores": 64, - "cpu_threads": 128, - "mem_total_gib": 125.36544799804688 - }, - "cpu_load": { - "avg_pct": 3.1, - "max_pct": 3.3, - "p95_pct": 3.2, - "samples": 70, - "status": "ok" - }, - "gpus": [ - { - "index": 0, - "uuid": "GPU-8281bc14-c076-e306-75b1-e91596be45e7", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:4B:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.255, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 15.036223085, - "samples": 15, - "avg_temp_c": 58.666666666666664, - "p95_temp_c": 64.3, - "avg_power_w": 102.08999999999999, - "p95_power_w": 124.32599999999992, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 18.77753269993211, - "temp_cv_pct": 5.9915574211200315, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 525.067614256, - "samples": 525, - "avg_temp_c": 71.64, - "p95_temp_c": 72, - "avg_power_w": 508.3619999999994, - "p95_power_w": 520.1840000000001, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 98.29523809523809, - "avg_mem_usage_pct": 38.43047619047619, - "clock_cv_pct": 0, - "power_cv_pct": 10.748058849985153, - "temp_cv_pct": 2.789715042814912, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.741717307, - "samples": 114, - "avg_temp_c": 49.62280701754386, - "p95_temp_c": 53, - "avg_power_w": 89.43131578947367, - "p95_power_w": 92.2615, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 1.4118633273458687, - "temp_cv_pct": 2.9165260445929433, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.9575, - "thermal_sustain_score": 100, - "stability_score": 78.50388230002969, - "interconnect_score": 0, - "composite_score": 530.3421839833501 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 1, - "uuid": "GPU-a66d6d68-9870-dfe0-1823-b38a141c21ae", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:4C:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.185, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 15.036223085, - "samples": 15, - "avg_temp_c": 60.333333333333336, - "p95_temp_c": 66.3, - "avg_power_w": 104.13399999999999, - "p95_power_w": 135.44299999999987, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 28.279218723013607, - "temp_cv_pct": 5.634275705627386, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 525.067614256, - "samples": 525, - "avg_temp_c": 74.02285714285715, - "p95_temp_c": 75, - "avg_power_w": 520.1055999999996, - "p95_power_w": 532.95, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 97.92761904761905, - "avg_mem_usage_pct": 38.944761904761904, - "clock_cv_pct": 0, - "power_cv_pct": 11.619346871453436, - "temp_cv_pct": 2.9818204514658326, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.741717307, - "samples": 114, - "avg_temp_c": 51.53508771929825, - "p95_temp_c": 55, - "avg_power_w": 88.07684210526322, - "p95_power_w": 90.61, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 1.3530652386403696, - "temp_cv_pct": 2.546658308018548, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.96916666666668, - "thermal_sustain_score": 100, - "stability_score": 76.76130625709312, - "interconnect_score": 0, - "composite_score": 527.9102110302674 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 2, - "uuid": "GPU-82b32f5b-5fca-9674-a845-cfd5da365d09", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:4E:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.255, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 15.036223085, - "samples": 15, - "avg_temp_c": 58.46666666666667, - "p95_temp_c": 64.3, - "avg_power_w": 104.23466666666667, - "p95_power_w": 128.58599999999993, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 20.35320302333862, - "temp_cv_pct": 5.986045495282411, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 525.067614256, - "samples": 525, - "avg_temp_c": 74.3047619047619, - "p95_temp_c": 75, - "avg_power_w": 530.595066666667, - "p95_power_w": 544.2040000000001, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 97.9047619047619, - "avg_mem_usage_pct": 38.54666666666667, - "clock_cv_pct": 0, - "power_cv_pct": 11.755075516052898, - "temp_cv_pct": 3.1451081246290715, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.741717307, - "samples": 114, - "avg_temp_c": 50.473684210526315, - "p95_temp_c": 54, - "avg_power_w": 91.29131578947388, - "p95_power_w": 94.4975, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 1.5813669058676583, - "temp_cv_pct": 2.9944744585943432, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.9575, - "thermal_sustain_score": 100, - "stability_score": 76.4898489678942, - "interconnect_score": 0, - "composite_score": 527.520023231423 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 3, - "uuid": "GPU-c88562a9-54cd-c3cf-c3fc-ef6c68ce5228", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:4F:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.265, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 15.036223085, - "samples": 15, - "avg_temp_c": 55.86666666666667, - "p95_temp_c": 61.3, - "avg_power_w": 95.958, - "p95_power_w": 116.39399999999993, - "avg_graphics_clock_mhz": 2425.2, - "p95_graphics_clock_mhz": 2430, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0.16160249004012395, - "power_cv_pct": 18.28206144786019, - "temp_cv_pct": 5.540900604412363, - "clock_drift_pct": 0.21971985718209908 - }, - "steady": { - "duration_sec": 525.067614256, - "samples": 525, - "avg_temp_c": 75.29142857142857, - "p95_temp_c": 76, - "avg_power_w": 525.9955047619045, - "p95_power_w": 539.4580000000001, - "avg_graphics_clock_mhz": 2422.0761904761903, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 97.94285714285714, - "avg_mem_usage_pct": 38.499047619047616, - "clock_cv_pct": 0.03207964525416391, - "power_cv_pct": 11.564343959222406, - "temp_cv_pct": 3.0543353777078432, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.741717307, - "samples": 114, - "avg_temp_c": 52.39473684210526, - "p95_temp_c": 55.349999999999994, - "avg_power_w": 88.51675438596483, - "p95_power_w": 91.02149999999999, - "avg_graphics_clock_mhz": 2422.2105263157896, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0.05286830237971597, - "power_cv_pct": 1.3571261625836293, - "temp_cv_pct": 2.5104636139944287, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.95583333333333, - "thermal_sustain_score": 100, - "stability_score": 76.74299350053853, - "interconnect_score": 0, - "composite_score": 527.8733403252803 - }, - "degradation_reasons": [ - "variance_too_high" - ] - } - ], - "server_power": { - "available": true, - "idle_w": 1048, - "loaded_w": 3737.5384615384614, - "delta_w": 2689.5384615384614, - "gpu_reported_sum_w": 2085.0581714285704, - "reporting_ratio": 1.2899105158757913 - } -} \ No newline at end of file diff --git a/gpu-benchmark-20260413-050458/result.json b/gpu-benchmark-20260413-050458/result.json deleted file mode 100644 index aa41f05..0000000 --- a/gpu-benchmark-20260413-050458/result.json +++ /dev/null @@ -1,776 +0,0 @@ -{ - "benchmark_version": "1", - "generated_at": "2026-04-13T05:04:58.909534574Z", - "hostname": "debian", - "server_model": "G5500 V7", - "benchmark_profile": "standard", - "parallel_gpus": true, - "ramp_step": 5, - "ramp_total": 8, - "ramp_run_id": "ramp-20260413-035309", - "overall_status": "OK", - "selected_gpu_indices": [ - 0, - 1, - 2, - 3, - 4 - ], - "findings": [ - "All 5 GPU(s) passed the benchmark.", - "GPU 0 showed unstable clocks/power over the benchmark window.", - "GPU 1 showed unstable clocks/power over the benchmark window.", - "GPU 2 showed unstable clocks/power over the benchmark window.", - "GPU 3 showed unstable clocks/power over the benchmark window.", - "GPU 4 showed unstable clocks/power over the benchmark window." - ], - "normalization": { - "status": "full", - "gpus": [ - { - "index": 0, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 1, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 2, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 3, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 4, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - } - ] - }, - "host_config": { - "cpu_model": "INTEL(R) XEON(R) GOLD 6530", - "cpu_sockets": 2, - "cpu_cores": 64, - "cpu_threads": 128, - "mem_total_gib": 125.36544799804688 - }, - "cpu_load": { - "avg_pct": 3.7, - "max_pct": 3.9, - "p95_pct": 3.9, - "samples": 70, - "status": "ok" - }, - "gpus": [ - { - "index": 0, - "uuid": "GPU-8281bc14-c076-e306-75b1-e91596be45e7", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:4B:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.2645, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 15.003491683, - "samples": 15, - "avg_temp_c": 58.8, - "p95_temp_c": 64.3, - "avg_power_w": 97.76533333333332, - "p95_power_w": 104.70400000000001, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 3.7803015697327367, - "temp_cv_pct": 5.798962291403367, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 527.066986904, - "samples": 527, - "avg_temp_c": 71.57495256166983, - "p95_temp_c": 73, - "avg_power_w": 507.0497533206832, - "p95_power_w": 520.9, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 97.74952561669829, - "avg_mem_usage_pct": 37.86337760910816, - "clock_cv_pct": 0, - "power_cv_pct": 12.012757746133422, - "temp_cv_pct": 3.2586034768067687, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.648768617, - "samples": 114, - "avg_temp_c": 49.56140350877193, - "p95_temp_c": 52, - "avg_power_w": 89.30096491228069, - "p95_power_w": 91.444, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 1.1038088274923465, - "temp_cv_pct": 2.4499053413094867, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.95591666666667, - "thermal_sustain_score": 100, - "stability_score": 75.97448450773315, - "interconnect_score": 0, - "composite_score": 526.796538479771 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 1, - "uuid": "GPU-a66d6d68-9870-dfe0-1823-b38a141c21ae", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:4C:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.249, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 15.003491683, - "samples": 15, - "avg_temp_c": 60.93333333333333, - "p95_temp_c": 66.3, - "avg_power_w": 96.53799999999998, - "p95_power_w": 102.759, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 3.5395358534781804, - "temp_cv_pct": 5.441938781376757, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 527.066986904, - "samples": 527, - "avg_temp_c": 74.3965844402277, - "p95_temp_c": 75, - "avg_power_w": 518.9630170777989, - "p95_power_w": 533.52, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 97.55977229601518, - "avg_mem_usage_pct": 38.588235294117645, - "clock_cv_pct": 0, - "power_cv_pct": 12.701708089948635, - "temp_cv_pct": 3.4531352747066717, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.648768617, - "samples": 114, - "avg_temp_c": 51.8421052631579, - "p95_temp_c": 54, - "avg_power_w": 88.07710526315786, - "p95_power_w": 90.59700000000001, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 1.2064879357740999, - "temp_cv_pct": 2.115803591518185, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.9585, - "thermal_sustain_score": 100, - "stability_score": 74.59658382010272, - "interconnect_score": 0, - "composite_score": 524.8679294396218 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 2, - "uuid": "GPU-82b32f5b-5fca-9674-a845-cfd5da365d09", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:4E:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.2090000000001, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 15.003491683, - "samples": 15, - "avg_temp_c": 58.86666666666667, - "p95_temp_c": 64.3, - "avg_power_w": 101.12600000000002, - "p95_power_w": 109.22299999999998, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 4.773882905476738, - "temp_cv_pct": 5.646636857222156, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 527.066986904, - "samples": 527, - "avg_temp_c": 74.14800759013282, - "p95_temp_c": 75, - "avg_power_w": 528.8367552182166, - "p95_power_w": 544.247, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 97.5370018975332, - "avg_mem_usage_pct": 38.415559772296014, - "clock_cv_pct": 0, - "power_cv_pct": 12.766621799252786, - "temp_cv_pct": 3.59396235562364, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.648768617, - "samples": 114, - "avg_temp_c": 50.526315789473685, - "p95_temp_c": 53.349999999999994, - "avg_power_w": 91.43912280701771, - "p95_power_w": 94.1305, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 1.2921924053692504, - "temp_cv_pct": 2.44810135237115, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.96516666666666, - "thermal_sustain_score": 100, - "stability_score": 74.46675640149442, - "interconnect_score": 0, - "composite_score": 524.6916139851322 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 3, - "uuid": "GPU-c88562a9-54cd-c3cf-c3fc-ef6c68ce5228", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:4F:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.269, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 15.003491683, - "samples": 15, - "avg_temp_c": 60.93333333333333, - "p95_temp_c": 66.3, - "avg_power_w": 96.50600000000001, - "p95_power_w": 104.157, - "avg_graphics_clock_mhz": 2425.2, - "p95_graphics_clock_mhz": 2430, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0.16160249004012395, - "power_cv_pct": 3.9969044026420764, - "temp_cv_pct": 5.441938781376757, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 527.066986904, - "samples": 527, - "avg_temp_c": 75.23908918406072, - "p95_temp_c": 76, - "avg_power_w": 524.9754079696397, - "p95_power_w": 540.1080000000001, - "avg_graphics_clock_mhz": 2422.0607210626185, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 97.66223908918406, - "avg_mem_usage_pct": 38.6584440227704, - "clock_cv_pct": 0.02866653390270464, - "power_cv_pct": 12.770181756684392, - "temp_cv_pct": 3.444650082421283, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.648768617, - "samples": 114, - "avg_temp_c": 52.44736842105263, - "p95_temp_c": 55, - "avg_power_w": 88.59570175438584, - "p95_power_w": 90.821, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 1.1068190849003288, - "temp_cv_pct": 2.1442860694231825, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.95516666666667, - "thermal_sustain_score": 100, - "stability_score": 74.3449703510204, - "interconnect_score": 0, - "composite_score": 524.5125540011004 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 4, - "uuid": "GPU-19d64ace-f127-d9a8-1999-ffe37453b930", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:CB:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.3885, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 15.003491683, - "samples": 15, - "avg_temp_c": 51.13333333333333, - "p95_temp_c": 56.3, - "avg_power_w": 88.92800000000003, - "p95_power_w": 94.11599999999999, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 3.1242071347047644, - "temp_cv_pct": 6.0538131766591405, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 527.066986904, - "samples": 527, - "avg_temp_c": 68.54459203036053, - "p95_temp_c": 69, - "avg_power_w": 510.0916698292221, - "p95_power_w": 523.961, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 97.91271347248576, - "avg_mem_usage_pct": 37.77988614800759, - "clock_cv_pct": 0, - "power_cv_pct": 12.20243417529326, - "temp_cv_pct": 3.207085724241198, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.648768617, - "samples": 114, - "avg_temp_c": 46.578947368421055, - "p95_temp_c": 50, - "avg_power_w": 85.92385964912292, - "p95_power_w": 88.47, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 1.1661917565791895, - "temp_cv_pct": 2.8844929564663992, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.93525, - "thermal_sustain_score": 100, - "stability_score": 75.59513164941347, - "interconnect_score": 0, - "composite_score": 526.2475954590352 - }, - "degradation_reasons": [ - "variance_too_high" - ] - } - ], - "server_power": { - "available": true, - "idle_w": 1181.3333333333333, - "loaded_w": 4268.7692307692305, - "delta_w": 3087.4358974358975, - "gpu_reported_sum_w": 2589.9166034155605, - "reporting_ratio": 1.1920985769828312 - } -} \ No newline at end of file diff --git a/gpu-benchmark-20260413-052002/result.json b/gpu-benchmark-20260413-052002/result.json deleted file mode 100644 index 5ba822c..0000000 --- a/gpu-benchmark-20260413-052002/result.json +++ /dev/null @@ -1,922 +0,0 @@ -{ - "benchmark_version": "1", - "generated_at": "2026-04-13T05:20:02.176595191Z", - "hostname": "debian", - "server_model": "G5500 V7", - "benchmark_profile": "standard", - "parallel_gpus": true, - "ramp_step": 6, - "ramp_total": 8, - "ramp_run_id": "ramp-20260413-035309", - "overall_status": "OK", - "selected_gpu_indices": [ - 0, - 1, - 2, - 3, - 4, - 5 - ], - "findings": [ - "All 6 GPU(s) passed the benchmark.", - "GPU 0 showed unstable clocks/power over the benchmark window.", - "GPU 1 showed unstable clocks/power over the benchmark window.", - "GPU 2 showed unstable clocks/power over the benchmark window.", - "GPU 3 showed unstable clocks/power over the benchmark window.", - "GPU 4 showed unstable clocks/power over the benchmark window.", - "GPU 5 showed unstable clocks/power over the benchmark window." - ], - "normalization": { - "status": "full", - "gpus": [ - { - "index": 0, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 1, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 2, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 3, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 4, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 5, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - } - ] - }, - "host_config": { - "cpu_model": "INTEL(R) XEON(R) GOLD 6530", - "cpu_sockets": 2, - "cpu_cores": 64, - "cpu_threads": 128, - "mem_total_gib": 125.36544799804688 - }, - "cpu_load": { - "avg_pct": 3.8, - "max_pct": 4.9, - "p95_pct": 4.9, - "samples": 85, - "status": "ok" - }, - "gpus": [ - { - "index": 0, - "uuid": "GPU-8281bc14-c076-e306-75b1-e91596be45e7", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:4B:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.3, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 16.053059923, - "samples": 16, - "avg_temp_c": 58.625, - "p95_temp_c": 64.25, - "avg_power_w": 97.95062499999999, - "p95_power_w": 104.6425, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 3.5538544062758373, - "temp_cv_pct": 5.843151217929519, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 529.066805487, - "samples": 529, - "avg_temp_c": 73.03024574669188, - "p95_temp_c": 74, - "avg_power_w": 516.1044234404536, - "p95_power_w": 532.3919999999999, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 97.3742911153119, - "avg_mem_usage_pct": 38.68620037807183, - "clock_cv_pct": 0, - "power_cv_pct": 13.26954724794044, - "temp_cv_pct": 3.837795991456247, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.562160642, - "samples": 114, - "avg_temp_c": 49.473684210526315, - "p95_temp_c": 52.349999999999994, - "avg_power_w": 89.25877192982453, - "p95_power_w": 91.7115, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 1.123523691008624, - "temp_cv_pct": 2.2750375790560877, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.95, - "thermal_sustain_score": 100, - "stability_score": 73.46090550411913, - "interconnect_score": 0, - "composite_score": 523.2694157811586 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 1, - "uuid": "GPU-a66d6d68-9870-dfe0-1823-b38a141c21ae", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:4C:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.204, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 16.053059923, - "samples": 16, - "avg_temp_c": 60.875, - "p95_temp_c": 66.25, - "avg_power_w": 96.51062499999996, - "p95_power_w": 102.8975, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 3.6149766986050422, - "temp_cv_pct": 5.287225129192498, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 529.066805487, - "samples": 529, - "avg_temp_c": 74.37240075614366, - "p95_temp_c": 75, - "avg_power_w": 517.2094328922498, - "p95_power_w": 533.688, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 97.19281663516068, - "avg_mem_usage_pct": 37.65217391304348, - "clock_cv_pct": 0, - "power_cv_pct": 13.84280954041754, - "temp_cv_pct": 3.6879251320846564, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.562160642, - "samples": 114, - "avg_temp_c": 51.3421052631579, - "p95_temp_c": 54, - "avg_power_w": 88.02447368421053, - "p95_power_w": 90.24149999999999, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 1.1123685256067952, - "temp_cv_pct": 1.8392656383518242, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.96600000000001, - "thermal_sustain_score": 100, - "stability_score": 72.31438091916492, - "interconnect_score": 0, - "composite_score": 521.6763021566902 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 2, - "uuid": "GPU-82b32f5b-5fca-9674-a845-cfd5da365d09", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:4E:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.238, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 16.053059923, - "samples": 16, - "avg_temp_c": 58.625, - "p95_temp_c": 64.25, - "avg_power_w": 99.44687500000002, - "p95_power_w": 107.8875, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 4.262486998727176, - "temp_cv_pct": 5.811945914899365, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 529.066805487, - "samples": 529, - "avg_temp_c": 74.24385633270322, - "p95_temp_c": 75, - "avg_power_w": 527.5048582230622, - "p95_power_w": 544.214, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 97.19092627599244, - "avg_mem_usage_pct": 37.54820415879017, - "clock_cv_pct": 0, - "power_cv_pct": 13.666807767183627, - "temp_cv_pct": 3.857745887762385, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.562160642, - "samples": 114, - "avg_temp_c": 50.37719298245614, - "p95_temp_c": 53, - "avg_power_w": 92.15017543859648, - "p95_power_w": 94.5575, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 1.1369648747132337, - "temp_cv_pct": 1.991810636420684, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.96033333333332, - "thermal_sustain_score": 100, - "stability_score": 72.66638446563275, - "interconnect_score": 0, - "composite_score": 522.1647822782088 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 3, - "uuid": "GPU-c88562a9-54cd-c3cf-c3fc-ef6c68ce5228", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:4F:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.284, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 16.053059923, - "samples": 16, - "avg_temp_c": 60.8125, - "p95_temp_c": 66.25, - "avg_power_w": 96.78375000000001, - "p95_power_w": 103.835, - "avg_graphics_clock_mhz": 2424.5, - "p95_graphics_clock_mhz": 2430, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0.1529428436192135, - "power_cv_pct": 4.0949357562590665, - "temp_cv_pct": 5.3036241921647145, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 529.066805487, - "samples": 529, - "avg_temp_c": 75.4366729678639, - "p95_temp_c": 76, - "avg_power_w": 523.9496975425328, - "p95_power_w": 540.62, - "avg_graphics_clock_mhz": 2422.0151228733457, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 97.35349716446125, - "avg_mem_usage_pct": 38.52741020793951, - "clock_cv_pct": 0.014347440279947565, - "power_cv_pct": 13.778878893679746, - "temp_cv_pct": 3.609529116252534, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.562160642, - "samples": 114, - "avg_temp_c": 52.333333333333336, - "p95_temp_c": 54.349999999999994, - "avg_power_w": 88.46842105263153, - "p95_power_w": 90.41649999999998, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 1.007904824255818, - "temp_cv_pct": 1.8015459393287852, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.95266666666667, - "thermal_sustain_score": 100, - "stability_score": 72.38485245152071, - "interconnect_score": 0, - "composite_score": 521.7638402842446 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 4, - "uuid": "GPU-19d64ace-f127-d9a8-1999-ffe37453b930", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:CB:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.32, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 16.053059923, - "samples": 16, - "avg_temp_c": 54.875, - "p95_temp_c": 60.25, - "avg_power_w": 92.22937500000002, - "p95_power_w": 98.205, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 3.382386785564523, - "temp_cv_pct": 5.9006076751656575, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 529.066805487, - "samples": 529, - "avg_temp_c": 69.15879017013232, - "p95_temp_c": 70, - "avg_power_w": 509.86896030245765, - "p95_power_w": 525.116, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 97.53308128544424, - "avg_mem_usage_pct": 37.90359168241966, - "clock_cv_pct": 0, - "power_cv_pct": 13.0988749354967, - "temp_cv_pct": 3.6747327643850913, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.562160642, - "samples": 114, - "avg_temp_c": 46.5, - "p95_temp_c": 49.349999999999994, - "avg_power_w": 85.81798245614041, - "p95_power_w": 87.93700000000001, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 0.9711049131181364, - "temp_cv_pct": 2.503564953095039, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.94666666666666, - "thermal_sustain_score": 100, - "stability_score": 73.80225012900661, - "interconnect_score": 0, - "composite_score": 523.7449218546949 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 5, - "uuid": "GPU-8d5d3b52-6221-c572-e4dc-3eb34699d5a4", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:CC:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.272, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 16.053059923, - "samples": 16, - "avg_temp_c": 53.375, - "p95_temp_c": 58.25, - "avg_power_w": 93.08749999999998, - "p95_power_w": 98.725, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 3.571040763944285, - "temp_cv_pct": 5.416877520477929, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 529.066805487, - "samples": 529, - "avg_temp_c": 71.84499054820417, - "p95_temp_c": 73, - "avg_power_w": 518.6966351606807, - "p95_power_w": 536.0600000000001, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 97.17958412098298, - "avg_mem_usage_pct": 38.36862003780718, - "clock_cv_pct": 0, - "power_cv_pct": 13.736936451172705, - "temp_cv_pct": 3.995613888278051, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.562160642, - "samples": 114, - "avg_temp_c": 49.01754385964912, - "p95_temp_c": 51, - "avg_power_w": 88.78377192982451, - "p95_power_w": 90.8905, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 0.9946861228149281, - "temp_cv_pct": 1.9293961305574223, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.95466666666665, - "thermal_sustain_score": 100, - "stability_score": 72.5261270976546, - "interconnect_score": 0, - "composite_score": 521.9634826377489 - }, - "degradation_reasons": [ - "variance_too_high" - ] - } - ], - "server_power": { - "available": true, - "idle_w": 1228, - "loaded_w": 4799.076923076923, - "delta_w": 3571.076923076923, - "gpu_reported_sum_w": 3113.334007561437, - "reporting_ratio": 1.1470266005522547 - } -} \ No newline at end of file diff --git a/gpu-benchmark-20260413-053511/result.json b/gpu-benchmark-20260413-053511/result.json deleted file mode 100644 index f232c1d..0000000 --- a/gpu-benchmark-20260413-053511/result.json +++ /dev/null @@ -1,1068 +0,0 @@ -{ - "benchmark_version": "1", - "generated_at": "2026-04-13T05:35:11.571022479Z", - "hostname": "debian", - "server_model": "G5500 V7", - "benchmark_profile": "standard", - "parallel_gpus": true, - "ramp_step": 7, - "ramp_total": 8, - "ramp_run_id": "ramp-20260413-035309", - "overall_status": "OK", - "selected_gpu_indices": [ - 0, - 1, - 2, - 3, - 4, - 5, - 6 - ], - "findings": [ - "All 7 GPU(s) passed the benchmark.", - "GPU 0 showed unstable clocks/power over the benchmark window.", - "GPU 1 showed unstable clocks/power over the benchmark window.", - "GPU 2 showed unstable clocks/power over the benchmark window.", - "GPU 3 showed unstable clocks/power over the benchmark window.", - "GPU 4 showed unstable clocks/power over the benchmark window.", - "GPU 5 showed unstable clocks/power over the benchmark window.", - "GPU 6 showed unstable clocks/power over the benchmark window." - ], - "normalization": { - "status": "full", - "gpus": [ - { - "index": 0, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 1, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 2, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 3, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 4, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 5, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 6, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - } - ] - }, - "host_config": { - "cpu_model": "INTEL(R) XEON(R) GOLD 6530", - "cpu_sockets": 2, - "cpu_cores": 64, - "cpu_threads": 128, - "mem_total_gib": 125.36544799804688 - }, - "cpu_load": { - "avg_pct": 4.4, - "max_pct": 5.7, - "p95_pct": 5.6, - "samples": 84, - "status": "ok" - }, - "gpus": [ - { - "index": 0, - "uuid": "GPU-8281bc14-c076-e306-75b1-e91596be45e7", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:4B:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.314, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 16.04678938, - "samples": 16, - "avg_temp_c": 58.1875, - "p95_temp_c": 64, - "avg_power_w": 97.07624999999999, - "p95_power_w": 104.1875, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 3.8062564565253245, - "temp_cv_pct": 5.641910198353984, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 531.065115947, - "samples": 531, - "avg_temp_c": 72.64783427495291, - "p95_temp_c": 74, - "avg_power_w": 510.3100188323918, - "p95_power_w": 532.09, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 96.99435028248588, - "avg_mem_usage_pct": 38.71939736346516, - "clock_cv_pct": 0, - "power_cv_pct": 14.24939934825413, - "temp_cv_pct": 4.027206872308068, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.372856967, - "samples": 114, - "avg_temp_c": 50.12280701754386, - "p95_temp_c": 52, - "avg_power_w": 89.38657894736835, - "p95_power_w": 91.3775, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 0.9290319187872187, - "temp_cv_pct": 1.8524492295973676, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.94766666666668, - "thermal_sustain_score": 100, - "stability_score": 71.50120130349174, - "interconnect_score": 0, - "composite_score": 520.5214218839666 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 1, - "uuid": "GPU-a66d6d68-9870-dfe0-1823-b38a141c21ae", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:4C:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.258, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 16.04678938, - "samples": 16, - "avg_temp_c": 60.5625, - "p95_temp_c": 66.25, - "avg_power_w": 95.88437499999998, - "p95_power_w": 102.965, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 3.962008587086827, - "temp_cv_pct": 5.748666450858805, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 531.065115947, - "samples": 531, - "avg_temp_c": 74.55743879472693, - "p95_temp_c": 76, - "avg_power_w": 516.4025988700567, - "p95_power_w": 534.14, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 96.92467043314501, - "avg_mem_usage_pct": 38.4030131826742, - "clock_cv_pct": 0, - "power_cv_pct": 14.717500783253215, - "temp_cv_pct": 3.946380183758861, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.372856967, - "samples": 114, - "avg_temp_c": 51.75438596491228, - "p95_temp_c": 54, - "avg_power_w": 88.00070175438614, - "p95_power_w": 90.0505, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 0.9245185916668872, - "temp_cv_pct": 1.7274837025411365, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.957, - "thermal_sustain_score": 100, - "stability_score": 70.56499843349357, - "interconnect_score": 0, - "composite_score": 519.2174162086708 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 2, - "uuid": "GPU-82b32f5b-5fca-9674-a845-cfd5da365d09", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:4E:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.2040000000001, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 16.04678938, - "samples": 16, - "avg_temp_c": 58.4375, - "p95_temp_c": 64.25, - "avg_power_w": 98.92562500000001, - "p95_power_w": 106.3175, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 3.941158294591091, - "temp_cv_pct": 6.079346092458744, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 531.065115947, - "samples": 531, - "avg_temp_c": 74.05461393596987, - "p95_temp_c": 75, - "avg_power_w": 525.6578342749533, - "p95_power_w": 544.575, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 96.76459510357816, - "avg_mem_usage_pct": 38.220338983050844, - "clock_cv_pct": 0, - "power_cv_pct": 14.662926314872996, - "temp_cv_pct": 4.151712016818929, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.372856967, - "samples": 114, - "avg_temp_c": 50.40350877192982, - "p95_temp_c": 53, - "avg_power_w": 91.19526315789464, - "p95_power_w": 93.51750000000001, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 1.0123803885891556, - "temp_cv_pct": 1.9964678593386713, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.966, - "thermal_sustain_score": 100, - "stability_score": 70.674147370254, - "interconnect_score": 0, - "composite_score": 519.3779277098173 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 3, - "uuid": "GPU-c88562a9-54cd-c3cf-c3fc-ef6c68ce5228", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:4F:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.318, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 16.04678938, - "samples": 16, - "avg_temp_c": 60.5625, - "p95_temp_c": 66.25, - "avg_power_w": 95.94187500000001, - "p95_power_w": 102.9875, - "avg_graphics_clock_mhz": 2426, - "p95_graphics_clock_mhz": 2430, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0.16488046166529266, - "power_cv_pct": 3.9902501502275056, - "temp_cv_pct": 5.748666450858805, - "clock_drift_pct": 0.08237232289950577 - }, - "steady": { - "duration_sec": 531.065115947, - "samples": 531, - "avg_temp_c": 75.23728813559322, - "p95_temp_c": 76, - "avg_power_w": 521.9880979284374, - "p95_power_w": 540.245, - "avg_graphics_clock_mhz": 2422.030131826742, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 97.00376647834275, - "avg_mem_usage_pct": 38.22975517890772, - "clock_cv_pct": 0.020232929020700553, - "power_cv_pct": 14.608025160871934, - "temp_cv_pct": 3.9179887726622487, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.372856967, - "samples": 114, - "avg_temp_c": 52.29824561403509, - "p95_temp_c": 54, - "avg_power_w": 88.52728070175428, - "p95_power_w": 90.384, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 0.8770139832436826, - "temp_cv_pct": 1.5597890556944174, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.947, - "thermal_sustain_score": 100, - "stability_score": 70.70301796217333, - "interconnect_score": 0, - "composite_score": 519.4024083467765 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 4, - "uuid": "GPU-19d64ace-f127-d9a8-1999-ffe37453b930", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:CB:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.3, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 16.04678938, - "samples": 16, - "avg_temp_c": 54.9375, - "p95_temp_c": 60.25, - "avg_power_w": 92.15750000000004, - "p95_power_w": 98.3625, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 3.352429847224058, - "temp_cv_pct": 5.879603895072832, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 531.065115947, - "samples": 531, - "avg_temp_c": 69.18455743879473, - "p95_temp_c": 70, - "avg_power_w": 508.81698681732604, - "p95_power_w": 525.855, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 97.14124293785311, - "avg_mem_usage_pct": 38.2090395480226, - "clock_cv_pct": 0, - "power_cv_pct": 14.121257794447253, - "temp_cv_pct": 4.066965153480223, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.372856967, - "samples": 114, - "avg_temp_c": 47.06140350877193, - "p95_temp_c": 49, - "avg_power_w": 85.78964912280702, - "p95_power_w": 87.43, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 0.8815489349736993, - "temp_cv_pct": 2.0351040458762837, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.95, - "thermal_sustain_score": 100, - "stability_score": 71.7574844111055, - "interconnect_score": 0, - "composite_score": 520.8824998964992 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 5, - "uuid": "GPU-8d5d3b52-6221-c572-e4dc-3eb34699d5a4", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:CC:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.264, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 16.04678938, - "samples": 16, - "avg_temp_c": 57.5, - "p95_temp_c": 63.25, - "avg_power_w": 96.57249999999999, - "p95_power_w": 103.7375, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 3.902165660400851, - "temp_cv_pct": 6.086956521739131, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 531.065115947, - "samples": 531, - "avg_temp_c": 71.94161958568738, - "p95_temp_c": 73, - "avg_power_w": 516.3799811676082, - "p95_power_w": 536.86, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 96.8060263653484, - "avg_mem_usage_pct": 38.133709981167605, - "clock_cv_pct": 0, - "power_cv_pct": 14.786168761467863, - "temp_cv_pct": 4.231629191129241, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.372856967, - "samples": 114, - "avg_temp_c": 49.280701754385966, - "p95_temp_c": 51, - "avg_power_w": 88.84307017543861, - "p95_power_w": 90.644, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 0.942977011916492, - "temp_cv_pct": 1.5788847890022546, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.956, - "thermal_sustain_score": 100, - "stability_score": 70.42766247706427, - "interconnect_score": 0, - "composite_score": 519.0241336863296 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 6, - "uuid": "GPU-a23c67ec-a8e2-7ebf-2593-79d73ec889a6", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:CE:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.25, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 16.04678938, - "samples": 16, - "avg_temp_c": 51.5, - "p95_temp_c": 57, - "avg_power_w": 92.36124999999998, - "p95_power_w": 97.38, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 2.950204809082421, - "temp_cv_pct": 5.905594689609921, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 531.065115947, - "samples": 531, - "avg_temp_c": 70.20338983050847, - "p95_temp_c": 71, - "avg_power_w": 509.8354237288135, - "p95_power_w": 527.455, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 96.84934086629002, - "avg_mem_usage_pct": 38.18079096045198, - "clock_cv_pct": 0, - "power_cv_pct": 14.468544117030715, - "temp_cv_pct": 4.251349242502187, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.372856967, - "samples": 114, - "avg_temp_c": 47.35087719298246, - "p95_temp_c": 49.349999999999994, - "avg_power_w": 87.85578947368428, - "p95_power_w": 89.6515, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 0.983368735998178, - "temp_cv_pct": 1.977622251957387, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.95833333333333, - "thermal_sustain_score": 100, - "stability_score": 71.06291176593857, - "interconnect_score": 0, - "composite_score": 519.9162374103023 - }, - "degradation_reasons": [ - "variance_too_high" - ] - } - ], - "server_power": { - "available": true, - "idle_w": 1164, - "loaded_w": 5288.114285714286, - "delta_w": 4124.114285714286, - "gpu_reported_sum_w": 3609.3909416195866, - "reporting_ratio": 1.1426067035741314 - } -} \ No newline at end of file diff --git a/gpu-benchmark-20260413-055026/result.json b/gpu-benchmark-20260413-055026/result.json deleted file mode 100644 index 2101463..0000000 --- a/gpu-benchmark-20260413-055026/result.json +++ /dev/null @@ -1,1234 +0,0 @@ -{ - "benchmark_version": "1", - "generated_at": "2026-04-13T05:50:26.607897085Z", - "hostname": "debian", - "server_model": "G5500 V7", - "benchmark_profile": "standard", - "parallel_gpus": true, - "ramp_step": 8, - "ramp_total": 8, - "ramp_run_id": "ramp-20260413-035309", - "overall_status": "OK", - "selected_gpu_indices": [ - 0, - 1, - 2, - 3, - 4, - 5, - 6, - 7 - ], - "findings": [ - "All 8 GPU(s) passed the benchmark.", - "GPU 0 showed unstable clocks/power over the benchmark window.", - "GPU 1 showed unstable clocks/power over the benchmark window.", - "GPU 2 showed unstable clocks/power over the benchmark window.", - "GPU 3 showed unstable clocks/power over the benchmark window.", - "GPU 4 showed unstable clocks/power over the benchmark window.", - "GPU 5 showed unstable clocks/power over the benchmark window.", - "GPU 6 showed unstable clocks/power over the benchmark window.", - "GPU 7 showed unstable clocks/power over the benchmark window.", - "Multi-GPU all_reduce max bus bandwidth: 3.9 GB/s." - ], - "normalization": { - "status": "full", - "gpus": [ - { - "index": 0, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 1, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 2, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 3, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 4, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 5, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 6, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - }, - { - "index": 7, - "persistence_mode": "applied", - "gpu_clock_lock_mhz": 2430, - "gpu_clock_lock_status": "applied", - "memory_clock_lock_mhz": 12481, - "memory_clock_lock_status": "applied" - } - ] - }, - "host_config": { - "cpu_model": "INTEL(R) XEON(R) GOLD 6530", - "cpu_sockets": 2, - "cpu_cores": 64, - "cpu_threads": 128, - "mem_total_gib": 125.36544799804688 - }, - "cpu_load": { - "avg_pct": 4.9, - "max_pct": 6.9, - "p95_pct": 6.6, - "samples": 91, - "status": "ok" - }, - "gpus": [ - { - "index": 0, - "uuid": "GPU-8281bc14-c076-e306-75b1-e91596be45e7", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:4B:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.284, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 16.034820522, - "samples": 16, - "avg_temp_c": 58.5625, - "p95_temp_c": 64.25, - "avg_power_w": 97.34562499999998, - "p95_power_w": 104.3425, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 3.8153775648467336, - "temp_cv_pct": 5.852307308796241, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 534.065632494, - "samples": 534, - "avg_temp_c": 72.97940074906367, - "p95_temp_c": 74, - "avg_power_w": 512.3819850187268, - "p95_power_w": 533.2760000000001, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 96.45692883895131, - "avg_mem_usage_pct": 38.39138576779026, - "clock_cv_pct": 0, - "power_cv_pct": 15.592428960695315, - "temp_cv_pct": 4.647413618844314, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.311993894, - "samples": 114, - "avg_temp_c": 50.19298245614035, - "p95_temp_c": 52, - "avg_power_w": 89.36491228070172, - "p95_power_w": 91.181, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 0.8317003174992245, - "temp_cv_pct": 1.1765139936880011, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.95266666666667, - "thermal_sustain_score": 100, - "stability_score": 68.81514207860937, - "interconnect_score": 3.92, - "composite_score": 572.8117211343068 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 1, - "uuid": "GPU-a66d6d68-9870-dfe0-1823-b38a141c21ae", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:4C:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.224, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 16.034820522, - "samples": 16, - "avg_temp_c": 60.625, - "p95_temp_c": 66.25, - "avg_power_w": 96.18749999999999, - "p95_power_w": 102.9575, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 3.6190678467089095, - "temp_cv_pct": 5.650387466410194, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 534.065632494, - "samples": 534, - "avg_temp_c": 74.33707865168539, - "p95_temp_c": 76, - "avg_power_w": 513.9588576779026, - "p95_power_w": 534.706, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 96.09550561797752, - "avg_mem_usage_pct": 37.97752808988764, - "clock_cv_pct": 0, - "power_cv_pct": 16.032037069751265, - "temp_cv_pct": 4.56941886281622, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.311993894, - "samples": 114, - "avg_temp_c": 51.333333333333336, - "p95_temp_c": 53, - "avg_power_w": 88.73859649122802, - "p95_power_w": 90.269, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 0.8602569271127033, - "temp_cv_pct": 1.4288820888835887, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.96266666666666, - "thermal_sustain_score": 100, - "stability_score": 67.93592586049746, - "interconnect_score": 3.92, - "composite_score": 571.5881284065297 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 2, - "uuid": "GPU-82b32f5b-5fca-9674-a845-cfd5da365d09", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:4E:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.1980000000001, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 16.034820522, - "samples": 16, - "avg_temp_c": 58.5625, - "p95_temp_c": 64.25, - "avg_power_w": 99.20875000000002, - "p95_power_w": 105.82, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 3.747848083593493, - "temp_cv_pct": 5.944992306170952, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 534.065632494, - "samples": 534, - "avg_temp_c": 74.38389513108615, - "p95_temp_c": 76, - "avg_power_w": 524.240243445693, - "p95_power_w": 545.7495, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 96.23970037453184, - "avg_mem_usage_pct": 38.12546816479401, - "clock_cv_pct": 0, - "power_cv_pct": 15.99414543851988, - "temp_cv_pct": 4.860573845224712, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.311993894, - "samples": 114, - "avg_temp_c": 50.333333333333336, - "p95_temp_c": 52, - "avg_power_w": 91.16175438596487, - "p95_power_w": 92.95, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 0.9826901628083612, - "temp_cv_pct": 1.5934652423370474, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.96699999999998, - "thermal_sustain_score": 100, - "stability_score": 68.01170912296024, - "interconnect_score": 3.92, - "composite_score": 571.6979628186027 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 3, - "uuid": "GPU-c88562a9-54cd-c3cf-c3fc-ef6c68ce5228", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:4F:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.3299999999999, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 16.034820522, - "samples": 16, - "avg_temp_c": 60.625, - "p95_temp_c": 66.25, - "avg_power_w": 96.16687500000002, - "p95_power_w": 103.22749999999999, - "avg_graphics_clock_mhz": 2426, - "p95_graphics_clock_mhz": 2430, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0.16488046166529266, - "power_cv_pct": 3.782241291247955, - "temp_cv_pct": 5.650387466410194, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 534.065632494, - "samples": 534, - "avg_temp_c": 75.62921348314607, - "p95_temp_c": 77, - "avg_power_w": 520.5265355805242, - "p95_power_w": 541.687, - "avg_graphics_clock_mhz": 2422.0898876404494, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 96.28651685393258, - "avg_mem_usage_pct": 37.99250936329588, - "clock_cv_pct": 0.03481377033352888, - "power_cv_pct": 15.961220894100176, - "temp_cv_pct": 4.648694191504124, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.311993894, - "samples": 114, - "avg_temp_c": 52.271929824561404, - "p95_temp_c": 54, - "avg_power_w": 88.49894736842091, - "p95_power_w": 89.7775, - "avg_graphics_clock_mhz": 2422.1403508771928, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0.04336205341685846, - "power_cv_pct": 0.8186267515664026, - "temp_cv_pct": 1.3476366177488701, - "clock_drift_pct": 0.011795234725168349 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.94500000000001, - "thermal_sustain_score": 100, - "stability_score": 67.93830313046553, - "interconnect_score": 3.92, - "composite_score": 571.5766063201771 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 4, - "uuid": "GPU-19d64ace-f127-d9a8-1999-ffe37453b930", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:CB:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.34, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 16.034820522, - "samples": 16, - "avg_temp_c": 55, - "p95_temp_c": 61, - "avg_power_w": 92.33500000000002, - "p95_power_w": 98.7525, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 3.492341021406794, - "temp_cv_pct": 6.064392756421061, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 534.065632494, - "samples": 534, - "avg_temp_c": 69.26779026217228, - "p95_temp_c": 71, - "avg_power_w": 506.3082958801495, - "p95_power_w": 526.117, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 96.45318352059925, - "avg_mem_usage_pct": 37.60299625468165, - "clock_cv_pct": 0, - "power_cv_pct": 15.54102422434703, - "temp_cv_pct": 4.641054486371343, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.311993894, - "samples": 114, - "avg_temp_c": 46.921052631578945, - "p95_temp_c": 49, - "avg_power_w": 86.75456140350886, - "p95_power_w": 88.41, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 0.7715872996388503, - "temp_cv_pct": 1.8107227998901132, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.94333333333333, - "thermal_sustain_score": 100, - "stability_score": 68.91795155130595, - "interconnect_score": 3.92, - "composite_score": 572.9479357411767 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 5, - "uuid": "GPU-8d5d3b52-6221-c572-e4dc-3eb34699d5a4", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:CC:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.188, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 16.034820522, - "samples": 16, - "avg_temp_c": 57.75, - "p95_temp_c": 63.25, - "avg_power_w": 97.051875, - "p95_power_w": 104.14, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 3.9002521641511976, - "temp_cv_pct": 5.79181305638946, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 534.065632494, - "samples": 534, - "avg_temp_c": 72.08988764044943, - "p95_temp_c": 74, - "avg_power_w": 514.1286891385772, - "p95_power_w": 537.069, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 96.2940074906367, - "avg_mem_usage_pct": 37.838951310861425, - "clock_cv_pct": 0, - "power_cv_pct": 15.864665152673057, - "temp_cv_pct": 4.671288336933857, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.311993894, - "samples": 114, - "avg_temp_c": 49.58771929824562, - "p95_temp_c": 51, - "avg_power_w": 89.05771929824559, - "p95_power_w": 91.04, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 0.8327686073487044, - "temp_cv_pct": 1.4826689027277737, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.96866666666666, - "thermal_sustain_score": 100, - "stability_score": 68.27066969465389, - "interconnect_score": 3.92, - "composite_score": 572.0622321237707 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 6, - "uuid": "GPU-a23c67ec-a8e2-7ebf-2593-79d73ec889a6", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:CE:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.2180000000001, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 16.034820522, - "samples": 16, - "avg_temp_c": 55.6875, - "p95_temp_c": 61.25, - "avg_power_w": 95.11187500000003, - "p95_power_w": 102.2825, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 3.814204476943954, - "temp_cv_pct": 6.013660302859176, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 534.065632494, - "samples": 534, - "avg_temp_c": 70.8314606741573, - "p95_temp_c": 72, - "avg_power_w": 508.88904494382024, - "p95_power_w": 529.0815, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 96.2621722846442, - "avg_mem_usage_pct": 37.840823970037455, - "clock_cv_pct": 0, - "power_cv_pct": 15.79881616328304, - "temp_cv_pct": 4.810144110611115, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.311993894, - "samples": 114, - "avg_temp_c": 48.03508771929825, - "p95_temp_c": 50, - "avg_power_w": 88.44701754385963, - "p95_power_w": 89.97, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 0.7855280810040005, - "temp_cv_pct": 1.6756858651503985, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.96366666666665, - "thermal_sustain_score": 100, - "stability_score": 68.40236767343391, - "interconnect_score": 3.92, - "composite_score": 572.242569945777 - }, - "degradation_reasons": [ - "variance_too_high" - ] - }, - { - "index": 7, - "uuid": "GPU-f9d537b7-86a1-407c-c137-3bc30fb8c2e8", - "name": "NVIDIA RTX PRO 6000 Blackwell Server Edition", - "bus_id": "00000000:CF:00.0", - "vbios": "98.02.67.00.0A", - "compute_capability": "12.0", - "backend": "cublasLt", - "status": "OK", - "power_limit_w": 600, - "default_power_limit_w": 600, - "calibrated_peak_power_w": 600.354, - "max_graphics_clock_mhz": 2430, - "max_memory_clock_mhz": 12481, - "locked_graphics_clock_mhz": 2430, - "locked_memory_clock_mhz": 12481, - "baseline": { - "duration_sec": 16.034820522, - "samples": 16, - "avg_temp_c": 54.1875, - "p95_temp_c": 59.25, - "avg_power_w": 90.46875000000001, - "p95_power_w": 95.4, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 3.169928895921586, - "temp_cv_pct": 5.544736028476085, - "clock_drift_pct": 0 - }, - "steady": { - "duration_sec": 534.065632494, - "samples": 534, - "avg_temp_c": 72.82771535580524, - "p95_temp_c": 74, - "avg_power_w": 508.06207865168585, - "p95_power_w": 527.48, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 96.44194756554307, - "avg_mem_usage_pct": 37.62734082397004, - "clock_cv_pct": 0, - "power_cv_pct": 15.469757701038384, - "temp_cv_pct": 4.561873839006074, - "clock_drift_pct": 0 - }, - "cooldown": { - "duration_sec": 120.311993894, - "samples": 114, - "avg_temp_c": 50.28947368421053, - "p95_temp_c": 52, - "avg_power_w": 86.8166666666667, - "p95_power_w": 88.6715, - "avg_graphics_clock_mhz": 2422, - "p95_graphics_clock_mhz": 2422, - "avg_memory_clock_mhz": 12481, - "p95_memory_clock_mhz": 12481, - "avg_usage_pct": 0, - "avg_mem_usage_pct": 0, - "clock_cv_pct": 0, - "power_cv_pct": 0.956688279977931, - "temp_cv_pct": 1.4834649538474145, - "clock_drift_pct": 0 - }, - "throttle_counters": { - "sw_power_cap_us": 0, - "sw_thermal_slowdown_us": 0, - "sync_boost_us": 0, - "hw_thermal_slowdown_us": 0, - "hw_power_brake_slowdown_us": 0 - }, - "precision_results": [ - { - "name": "fp16_tensor", - "category": "fp16_bf16", - "supported": true, - "lanes": 16, - "m": 11904, - "n": 11904, - "k": 11904, - "iterations": 27840, - "teraops_per_sec": 195.675631386624 - }, - { - "name": "fp32_tf32", - "category": "fp32_tf32", - "supported": true, - "lanes": 16, - "m": 8320, - "n": 8320, - "k": 8320, - "iterations": 27840, - "teraops_per_sec": 66.807922688 - }, - { - "name": "fp64", - "category": "fp64", - "supported": false, - "notes": "unsupported" - }, - { - "name": "fp8_e4m3", - "category": "fp8", - "supported": true, - "lanes": 16, - "m": 13696, - "n": 13696, - "k": 13696, - "iterations": 27840, - "teraops_per_sec": 298.015759794176 - }, - { - "name": "fp8_e5m2", - "category": "fp8", - "supported": false, - "notes": "unsupported" - } - ], - "scores": { - "compute_score": 560.4993138688001, - "power_sustain_score": 99.94099999999999, - "thermal_sustain_score": 100, - "stability_score": 69.06048459792324, - "interconnect_score": 3.92, - "composite_score": 573.1456981806597 - }, - "degradation_reasons": [ - "variance_too_high" - ] - } - ], - "interconnect": { - "status": "OK", - "attempted": true, - "supported": true, - "selected_gpu_indices": [ - 0, - 1, - 2, - 3, - 4, - 5, - 6, - 7 - ], - "avg_algbw_gbps": 2.2100000000000004, - "max_algbw_gbps": 2.24, - "avg_busbw_gbps": 3.8699999999999997, - "max_busbw_gbps": 3.92 - }, - "server_power": { - "available": true, - "idle_w": 1305.3333333333333, - "loaded_w": 5845.942857142857, - "delta_w": 4540.609523809524, - "gpu_reported_sum_w": 4108.495730337079, - "reporting_ratio": 1.1051756705700635 - } -} \ No newline at end of file diff --git a/gpu-benchmark-20260414-152824/result.json b/gpu-benchmark-20260414-152824/result.json new file mode 100644 index 0000000..e6db3b7 --- /dev/null +++ b/gpu-benchmark-20260414-152824/result.json @@ -0,0 +1,141 @@ +{ + "benchmark_version": "1", + "generated_at": "2026-04-14T15:28:24.793730899Z", + "hostname": "debian", + "server_model": "MLT-S06", + "benchmark_profile": "standard", + "parallel_gpus": true, + "ramp_step": 1, + "ramp_total": 2, + "ramp_run_id": "ramp-20260414-152824", + "overall_status": "FAILED", + "selected_gpu_indices": [ + 0 + ], + "findings": [ + "0 of 1 GPU(s) passed the benchmark.", + "GPU 0 average SM clock stayed below the requested lock target." + ], + "normalization": { + "status": "full", + "gpus": [ + { + "index": 0, + "persistence_mode": "applied", + "gpu_clock_lock_mhz": 1980, + "gpu_clock_lock_status": "applied", + "memory_clock_lock_mhz": 2619, + "memory_clock_lock_status": "applied" + } + ] + }, + "host_config": { + "cpu_model": "Intel(R) Xeon(R) Gold 6430", + "cpu_sockets": 2, + "cpu_cores": 64, + "cpu_threads": 128, + "mem_total_gib": 62.53376007080078 + }, + "gpus": [ + { + "index": 0, + "uuid": "GPU-9ee0af22-3dda-6f5c-1a13-35c63f324216", + "name": "NVIDIA H100 80GB HBM3", + "bus_id": "00000000:49:00.0", + "vbios": "96.00.74.00.01", + "status": "UNSUPPORTED", + "power_limit_w": 700, + "default_power_limit_w": 700, + "max_graphics_clock_mhz": 1980, + "max_memory_clock_mhz": 2619, + "locked_graphics_clock_mhz": 1980, + "locked_memory_clock_mhz": 2619, + "baseline": { + "duration_sec": 0, + "samples": 0, + "avg_temp_c": 0, + "p95_temp_c": 0, + "avg_power_w": 0, + "p95_power_w": 0, + "avg_graphics_clock_mhz": 0, + "p95_graphics_clock_mhz": 0, + "avg_memory_clock_mhz": 0, + "p95_memory_clock_mhz": 0, + "avg_usage_pct": 0, + "avg_mem_usage_pct": 0, + "clock_cv_pct": 0, + "power_cv_pct": 0, + "temp_cv_pct": 0, + "clock_drift_pct": 0 + }, + "steady": { + "duration_sec": 0, + "samples": 0, + "avg_temp_c": 0, + "p95_temp_c": 0, + "avg_power_w": 0, + "p95_power_w": 0, + "avg_graphics_clock_mhz": 0, + "p95_graphics_clock_mhz": 0, + "avg_memory_clock_mhz": 0, + "p95_memory_clock_mhz": 0, + "avg_usage_pct": 0, + "avg_mem_usage_pct": 0, + "clock_cv_pct": 0, + "power_cv_pct": 0, + "temp_cv_pct": 0, + "clock_drift_pct": 0 + }, + "cooldown": { + "duration_sec": 0, + "samples": 0, + "avg_temp_c": 0, + "p95_temp_c": 0, + "avg_power_w": 0, + "p95_power_w": 0, + "avg_graphics_clock_mhz": 0, + "p95_graphics_clock_mhz": 0, + "avg_memory_clock_mhz": 0, + "p95_memory_clock_mhz": 0, + "avg_usage_pct": 0, + "avg_mem_usage_pct": 0, + "clock_cv_pct": 0, + "power_cv_pct": 0, + "temp_cv_pct": 0, + "clock_drift_pct": 0 + }, + "throttle_counters": { + "sw_power_cap_us": 0, + "sw_thermal_slowdown_us": 0, + "sync_boost_us": 0, + "hw_thermal_slowdown_us": 0, + "hw_power_brake_slowdown_us": 0 + }, + "ecc": { + "corrected": 0, + "uncorrected": 0 + }, + "scores": { + "compute_score": 0, + "power_sustain_score": 0, + "thermal_sustain_score": 100, + "stability_score": 100, + "interconnect_score": 0, + "composite_score": 0 + }, + "degradation_reasons": [ + "low_sm_clock_vs_target" + ], + "notes": [ + "parallel warmup failed: context canceled", + "precision plan failed: context canceled" + ] + } + ], + "server_power": { + "available": false, + "notes": [ + "IPMI power reading unavailable; server-side power characterization skipped" + ] + } +} \ No newline at end of file diff --git a/gpu-benchmark-20260414-152949/result.json b/gpu-benchmark-20260414-152949/result.json new file mode 100644 index 0000000..d8ba6af --- /dev/null +++ b/gpu-benchmark-20260414-152949/result.json @@ -0,0 +1,385 @@ +{ + "benchmark_version": "1", + "generated_at": "2026-04-14T15:29:49.813251958Z", + "hostname": "debian", + "server_model": "MLT-S06", + "benchmark_profile": "standard", + "parallel_gpus": true, + "ramp_step": 1, + "ramp_total": 2, + "ramp_run_id": "ramp-20260414-152949", + "overall_status": "OK", + "selected_gpu_indices": [ + 0 + ], + "findings": [ + "All 1 GPU(s) passed the benchmark.", + "GPU 0 average SM clock stayed below the requested lock target." + ], + "normalization": { + "status": "full", + "gpus": [ + { + "index": 0, + "persistence_mode": "applied", + "gpu_clock_lock_mhz": 1980, + "gpu_clock_lock_status": "applied", + "memory_clock_lock_mhz": 2619, + "memory_clock_lock_status": "applied" + } + ] + }, + "host_config": { + "cpu_model": "Intel(R) Xeon(R) Gold 6430", + "cpu_sockets": 2, + "cpu_cores": 64, + "cpu_threads": 128, + "mem_total_gib": 62.53376007080078 + }, + "cpu_load": { + "avg_pct": 0.7, + "max_pct": 1.1, + "p95_pct": 1, + "samples": 78, + "status": "ok" + }, + "cooling": { + "available": true, + "avg_fan_rpm": 6083.203125, + "notes": [ + "fan duty cycle unavailable on this host; RPM-only fan telemetry was collected" + ] + }, + "gpus": [ + { + "index": 0, + "uuid": "GPU-9ee0af22-3dda-6f5c-1a13-35c63f324216", + "name": "NVIDIA H100 80GB HBM3", + "bus_id": "00000000:49:00.0", + "vbios": "96.00.74.00.01", + "compute_capability": "9.0", + "backend": "cublasLt", + "status": "OK", + "power_limit_w": 700, + "default_power_limit_w": 700, + "calibrated_peak_power_w": 698.316, + "max_graphics_clock_mhz": 1980, + "max_memory_clock_mhz": 2619, + "locked_graphics_clock_mhz": 1980, + "locked_memory_clock_mhz": 2619, + "baseline": { + "duration_sec": 22.679193237, + "samples": 1, + "avg_temp_c": 66, + "p95_temp_c": 66, + "avg_power_w": 244.62, + "p95_power_w": 244.62, + "avg_graphics_clock_mhz": 1830, + "p95_graphics_clock_mhz": 1830, + "avg_memory_clock_mhz": 2619, + "p95_memory_clock_mhz": 2619, + "avg_usage_pct": 0, + "avg_mem_usage_pct": 0, + "clock_cv_pct": 0, + "power_cv_pct": 0, + "temp_cv_pct": 0, + "clock_drift_pct": 0 + }, + "steady": { + "duration_sec": 664.658712143, + "samples": 13, + "avg_temp_c": 84.23076923076923, + "p95_temp_c": 87, + "avg_power_w": 531.6661538461539, + "p95_power_w": 612.742, + "avg_graphics_clock_mhz": 1221.923076923077, + "p95_graphics_clock_mhz": 1830, + "avg_memory_clock_mhz": 2619, + "p95_memory_clock_mhz": 2619, + "avg_usage_pct": 76.92307692307692, + "avg_mem_usage_pct": 38.07692307692308, + "clock_cv_pct": 27.28679189404879, + "power_cv_pct": 25.428186879992428, + "temp_cv_pct": 5.571524764445532, + "clock_drift_pct": 0 + }, + "precision_steady": [ + { + "precision": "int8", + "steady": { + "duration_sec": 50.799623108, + "samples": 2, + "avg_temp_c": 73.5, + "p95_temp_c": 76.65, + "avg_power_w": 641.45, + "p95_power_w": 692.912, + "avg_graphics_clock_mhz": 1245, + "p95_graphics_clock_mhz": 1299, + "avg_memory_clock_mhz": 2619, + "p95_memory_clock_mhz": 2619, + "avg_usage_pct": 100, + "avg_mem_usage_pct": 80.5, + "clock_cv_pct": 4.819277108433735, + "power_cv_pct": 8.91417881362538, + "temp_cv_pct": 4.761904761904762, + "clock_drift_pct": 0 + }, + "teraops_per_sec": 1104.4978767691775, + "weighted_teraops_per_sec": 276.1244691922944, + "ecc": { + "corrected": 0, + "uncorrected": 0 + } + }, + { + "precision": "fp8", + "steady": { + "duration_sec": 119.043156359, + "samples": 3, + "avg_temp_c": 84.33333333333333, + "p95_temp_c": 86.8, + "avg_power_w": 681.4366666666666, + "p95_power_w": 699.73, + "avg_graphics_clock_mhz": 1145, + "p95_graphics_clock_mhz": 1170, + "avg_memory_clock_mhz": 2619, + "p95_memory_clock_mhz": 2619, + "avg_usage_pct": 100, + "avg_mem_usage_pct": 52, + "clock_cv_pct": 3.087802537932522, + "power_cv_pct": 3.7139603513146184, + "temp_cv_pct": 2.957831926303511, + "clock_drift_pct": 0 + }, + "teraops_per_sec": 1096.7958204099243, + "weighted_teraops_per_sec": 274.1989551024811, + "ecc": { + "corrected": 0, + "uncorrected": 0 + } + }, + { + "precision": "fp16", + "steady": { + "duration_sec": 179.687679749, + "samples": 2, + "avg_temp_c": 87, + "p95_temp_c": 87, + "avg_power_w": 612.785, + "p95_power_w": 613.1315, + "avg_graphics_clock_mhz": 1087.5, + "p95_graphics_clock_mhz": 1107.75, + "avg_memory_clock_mhz": 2619, + "p95_memory_clock_mhz": 2619, + "avg_usage_pct": 100, + "avg_mem_usage_pct": 40.5, + "clock_cv_pct": 2.0689655172413794, + "power_cv_pct": 0.06282790864658745, + "temp_cv_pct": 0, + "clock_drift_pct": 0 + }, + "teraops_per_sec": 702.1073356117333, + "weighted_teraops_per_sec": 351.05366780586667, + "ecc": { + "corrected": 0, + "uncorrected": 0 + } + }, + { + "precision": "fp32", + "steady": { + "duration_sec": 232.355201725, + "samples": 2, + "avg_temp_c": 87, + "p95_temp_c": 87, + "avg_power_w": 603.5899999999999, + "p95_power_w": 609.4129999999999, + "avg_graphics_clock_mhz": 1102.5, + "p95_graphics_clock_mhz": 1109.25, + "avg_memory_clock_mhz": 2619, + "p95_memory_clock_mhz": 2619, + "avg_usage_pct": 100, + "avg_mem_usage_pct": 40, + "clock_cv_pct": 0.6802721088435374, + "power_cv_pct": 1.071919680577871, + "temp_cv_pct": 0, + "clock_drift_pct": 0 + }, + "teraops_per_sec": 297.2914085265408, + "weighted_teraops_per_sec": 297.2914085265408, + "ecc": { + "corrected": 0, + "uncorrected": 0 + } + }, + { + "precision": "fp64", + "steady": { + "duration_sec": 282.807869037, + "samples": 2, + "avg_temp_c": 87, + "p95_temp_c": 87, + "avg_power_w": 617.8499999999999, + "p95_power_w": 620.3879999999999, + "avg_graphics_clock_mhz": 1117.5, + "p95_graphics_clock_mhz": 1124.25, + "avg_memory_clock_mhz": 2619, + "p95_memory_clock_mhz": 2619, + "avg_usage_pct": 100, + "avg_mem_usage_pct": 47, + "clock_cv_pct": 0.6711409395973155, + "power_cv_pct": 0.4564214615197853, + "temp_cv_pct": 0, + "clock_drift_pct": 0 + }, + "ecc": { + "corrected": 0, + "uncorrected": 0 + } + }, + { + "precision": "fp4", + "steady": { + "duration_sec": 355.407457812, + "samples": 3, + "avg_temp_c": 86.33333333333333, + "p95_temp_c": 86.9, + "avg_power_w": 598.8033333333333, + "p95_power_w": 604.278, + "avg_graphics_clock_mhz": 1020, + "p95_graphics_clock_mhz": 1033.5, + "avg_memory_clock_mhz": 2619, + "p95_memory_clock_mhz": 2619, + "avg_usage_pct": 100, + "avg_mem_usage_pct": 47, + "clock_cv_pct": 1.2007302660701853, + "power_cv_pct": 0.7587124023422941, + "temp_cv_pct": 0.5460284024606545, + "clock_drift_pct": 0 + }, + "ecc": { + "corrected": 0, + "uncorrected": 0 + } + } + ], + "cooldown": { + "duration_sec": 0, + "samples": 0, + "avg_temp_c": 0, + "p95_temp_c": 0, + "avg_power_w": 0, + "p95_power_w": 0, + "avg_graphics_clock_mhz": 0, + "p95_graphics_clock_mhz": 0, + "avg_memory_clock_mhz": 0, + "p95_memory_clock_mhz": 0, + "avg_usage_pct": 0, + "avg_mem_usage_pct": 0, + "clock_cv_pct": 0, + "power_cv_pct": 0, + "temp_cv_pct": 0, + "clock_drift_pct": 0 + }, + "throttle_counters": { + "sw_power_cap_us": 0, + "sw_thermal_slowdown_us": 0, + "sync_boost_us": 0, + "hw_thermal_slowdown_us": 0, + "hw_power_brake_slowdown_us": 0 + }, + "ecc": { + "corrected": 0, + "uncorrected": 0 + }, + "precision_results": [ + { + "name": "fp16_tensor", + "category": "fp16_bf16", + "supported": true, + "lanes": 16, + "m": 9856, + "n": 9856, + "k": 9856, + "iterations": 28032, + "teraops_per_sec": 178.92248028971008, + "weight": 0.5, + "weighted_teraops_per_sec": 89.46124014485504 + }, + { + "name": "fp32_tf32", + "category": "fp32_tf32", + "supported": true, + "lanes": 16, + "m": 6912, + "n": 6912, + "k": 6912, + "iterations": 28032, + "teraops_per_sec": 61.71262413963264, + "weight": 1, + "weighted_teraops_per_sec": 61.71262413963264 + }, + { + "name": "fp64", + "category": "fp64", + "supported": false, + "weight": 2, + "notes": "unsupported" + }, + { + "name": "fp8_e4m3", + "category": "fp8", + "supported": true, + "lanes": 16, + "m": 11392, + "n": 11392, + "k": 11392, + "iterations": 28032, + "teraops_per_sec": 276.28846547206143, + "weight": 0.25, + "weighted_teraops_per_sec": 69.07211636801536 + }, + { + "name": "fp8_e5m2", + "category": "fp8", + "supported": false, + "weight": 0.25, + "notes": "unsupported" + }, + { + "name": "int8_tensor", + "category": "int8", + "supported": true, + "lanes": 16, + "m": 8832, + "n": 8832, + "k": 8832, + "iterations": 28032, + "teraops_per_sec": 128.74785629405184, + "weight": 0.25, + "weighted_teraops_per_sec": 32.18696407351296 + } + ], + "scores": { + "compute_score": 1274.3983840449878, + "synthetic_score": 1198.668500627183, + "mixed_score": 252.432944726016, + "mixed_efficiency": 0.21059445926370363, + "power_sustain_score": 99.75942857142857, + "thermal_sustain_score": 100, + "stability_score": 86.6552008085787, + "interconnect_score": 0, + "composite_score": 1231.422031972879 + }, + "degradation_reasons": [ + "low_sm_clock_vs_target" + ] + } + ], + "server_power": { + "available": false, + "notes": [ + "IPMI power reading unavailable; server-side power characterization skipped" + ] + } +} \ No newline at end of file diff --git a/gpu-benchmark-20260414-154657/result.json b/gpu-benchmark-20260414-154657/result.json new file mode 100644 index 0000000..a731346 --- /dev/null +++ b/gpu-benchmark-20260414-154657/result.json @@ -0,0 +1,735 @@ +{ + "benchmark_version": "1", + "generated_at": "2026-04-14T15:46:57.112694192Z", + "hostname": "debian", + "server_model": "MLT-S06", + "benchmark_profile": "standard", + "parallel_gpus": true, + "ramp_step": 2, + "ramp_total": 2, + "ramp_run_id": "ramp-20260414-152949", + "overall_status": "OK", + "selected_gpu_indices": [ + 0, + 1 + ], + "findings": [ + "All 2 GPU(s) passed the benchmark.", + "GPU 0 average SM clock stayed below the requested lock target.", + "GPU 0 showed unstable clocks/power over the benchmark window.", + "GPU 1 average SM clock stayed below the requested lock target.", + "GPU 1 showed unstable clocks/power over the benchmark window.", + "Multi-GPU all_reduce max bus bandwidth: 12.1 GB/s." + ], + "normalization": { + "status": "full", + "gpus": [ + { + "index": 0, + "persistence_mode": "applied", + "gpu_clock_lock_mhz": 1980, + "gpu_clock_lock_status": "applied", + "memory_clock_lock_mhz": 2619, + "memory_clock_lock_status": "applied" + }, + { + "index": 1, + "persistence_mode": "applied", + "gpu_clock_lock_mhz": 1980, + "gpu_clock_lock_status": "applied", + "memory_clock_lock_mhz": 2619, + "memory_clock_lock_status": "applied" + } + ] + }, + "host_config": { + "cpu_model": "Intel(R) Xeon(R) Gold 6430", + "cpu_sockets": 2, + "cpu_cores": 64, + "cpu_threads": 128, + "mem_total_gib": 62.53376007080078 + }, + "cpu_load": { + "avg_pct": 1.5, + "max_pct": 1.9, + "p95_pct": 1.8, + "samples": 73, + "status": "ok" + }, + "cooling": { + "available": true, + "avg_fan_rpm": 6134.375, + "notes": [ + "fan duty cycle unavailable on this host; RPM-only fan telemetry was collected" + ] + }, + "gpus": [ + { + "index": 0, + "uuid": "GPU-9ee0af22-3dda-6f5c-1a13-35c63f324216", + "name": "NVIDIA H100 80GB HBM3", + "bus_id": "00000000:49:00.0", + "vbios": "96.00.74.00.01", + "compute_capability": "9.0", + "backend": "cublasLt", + "status": "OK", + "power_limit_w": 700, + "default_power_limit_w": 700, + "max_graphics_clock_mhz": 1980, + "max_memory_clock_mhz": 2619, + "locked_graphics_clock_mhz": 1980, + "locked_memory_clock_mhz": 2619, + "baseline": { + "duration_sec": 20.631671609, + "samples": 1, + "avg_temp_c": 68, + "p95_temp_c": 68, + "avg_power_w": 252.73, + "p95_power_w": 252.73, + "avg_graphics_clock_mhz": 1830, + "p95_graphics_clock_mhz": 1830, + "avg_memory_clock_mhz": 2619, + "p95_memory_clock_mhz": 2619, + "avg_usage_pct": 0, + "avg_mem_usage_pct": 0, + "clock_cv_pct": 0, + "power_cv_pct": 0, + "temp_cv_pct": 0, + "clock_drift_pct": 0 + }, + "steady": { + "duration_sec": 603.015967594, + "samples": 12, + "avg_temp_c": 86.33333333333333, + "p95_temp_c": 87, + "avg_power_w": 586.6108333333333, + "p95_power_w": 596.883, + "avg_graphics_clock_mhz": 1023.75, + "p95_graphics_clock_mhz": 1050, + "avg_memory_clock_mhz": 2619, + "p95_memory_clock_mhz": 2619, + "avg_usage_pct": 100, + "avg_mem_usage_pct": 47.75, + "clock_cv_pct": 1.8069167993832407, + "power_cv_pct": 1.1372750232456648, + "temp_cv_pct": 0.5460284024606545, + "clock_drift_pct": 0 + }, + "precision_steady": [ + { + "precision": "int8", + "steady": { + "duration_sec": 58.789805111, + "samples": 2, + "avg_temp_c": 78.5, + "p95_temp_c": 81.65, + "avg_power_w": 612.745, + "p95_power_w": 689.4024999999999, + "avg_graphics_clock_mhz": 1222.5, + "p95_graphics_clock_mhz": 1269.75, + "avg_memory_clock_mhz": 2619, + "p95_memory_clock_mhz": 2619, + "avg_usage_pct": 100, + "avg_mem_usage_pct": 79.5, + "clock_cv_pct": 4.294478527607362, + "power_cv_pct": 13.90056222408995, + "temp_cv_pct": 4.45859872611465, + "clock_drift_pct": 0 + }, + "teraops_per_sec": 1095.3120993509376, + "weighted_teraops_per_sec": 273.8280248377344, + "ecc": { + "corrected": 0, + "uncorrected": 0 + } + }, + { + "precision": "fp8", + "steady": { + "duration_sec": 107.153140143, + "samples": 2, + "avg_temp_c": 86.5, + "p95_temp_c": 86.95, + "avg_power_w": 655.985, + "p95_power_w": 694.4375, + "avg_graphics_clock_mhz": 1102.5, + "p95_graphics_clock_mhz": 1136.25, + "avg_memory_clock_mhz": 2619, + "p95_memory_clock_mhz": 2619, + "avg_usage_pct": 100, + "avg_mem_usage_pct": 56, + "clock_cv_pct": 3.4013605442176873, + "power_cv_pct": 6.51310624480743, + "temp_cv_pct": 0.5780346820809248, + "clock_drift_pct": 0 + }, + "teraops_per_sec": 1093.6418424935764, + "weighted_teraops_per_sec": 273.4104606233941, + "ecc": { + "corrected": 0, + "uncorrected": 0 + } + }, + { + "precision": "fp16", + "steady": { + "duration_sec": 162.145217446, + "samples": 3, + "avg_temp_c": 87, + "p95_temp_c": 87, + "avg_power_w": 593.8366666666667, + "p95_power_w": 596.7439999999999, + "avg_graphics_clock_mhz": 1065, + "p95_graphics_clock_mhz": 1090.5, + "avg_memory_clock_mhz": 2619, + "p95_memory_clock_mhz": 2619, + "avg_usage_pct": 100, + "avg_mem_usage_pct": 36.666666666666664, + "clock_cv_pct": 1.9918500878494296, + "power_cv_pct": 0.6401906405475871, + "temp_cv_pct": 0, + "clock_drift_pct": 0 + }, + "teraops_per_sec": 699.5542180276906, + "weighted_teraops_per_sec": 349.7771090138453, + "ecc": { + "corrected": 0, + "uncorrected": 0 + } + }, + { + "precision": "fp32", + "steady": { + "duration_sec": 221.67717378, + "samples": 3, + "avg_temp_c": 87, + "p95_temp_c": 87, + "avg_power_w": 595.0466666666666, + "p95_power_w": 600.4, + "avg_graphics_clock_mhz": 1100, + "p95_graphics_clock_mhz": 1108.5, + "avg_memory_clock_mhz": 2619, + "p95_memory_clock_mhz": 2619, + "avg_usage_pct": 100, + "avg_mem_usage_pct": 33.666666666666664, + "clock_cv_pct": 0.6428243465332251, + "power_cv_pct": 1.0242875350150111, + "temp_cv_pct": 0, + "clock_drift_pct": 0 + }, + "teraops_per_sec": 296.763047018496, + "weighted_teraops_per_sec": 296.763047018496, + "ecc": { + "corrected": 0, + "uncorrected": 0 + } + }, + { + "precision": "fp64", + "steady": { + "duration_sec": 283.877134755, + "samples": 3, + "avg_temp_c": 87, + "p95_temp_c": 87, + "avg_power_w": 599.5233333333334, + "p95_power_w": 615.675, + "avg_graphics_clock_mhz": 1090, + "p95_graphics_clock_mhz": 1095, + "avg_memory_clock_mhz": 2619, + "p95_memory_clock_mhz": 2619, + "avg_usage_pct": 100, + "avg_mem_usage_pct": 46.333333333333336, + "clock_cv_pct": 0.6487218176023373, + "power_cv_pct": 2.322932483924268, + "temp_cv_pct": 0, + "clock_drift_pct": 0 + }, + "ecc": { + "corrected": 0, + "uncorrected": 0 + } + }, + { + "precision": "fp4", + "steady": { + "duration_sec": 353.430169396, + "samples": 3, + "avg_temp_c": 87, + "p95_temp_c": 87, + "avg_power_w": 594.26, + "p95_power_w": 602.7589999999999, + "avg_graphics_clock_mhz": 1020, + "p95_graphics_clock_mhz": 1047, + "avg_memory_clock_mhz": 2619, + "p95_memory_clock_mhz": 2619, + "avg_usage_pct": 100, + "avg_mem_usage_pct": 48.666666666666664, + "clock_cv_pct": 2.4014605321403706, + "power_cv_pct": 1.1839597038195573, + "temp_cv_pct": 0, + "clock_drift_pct": 0 + }, + "ecc": { + "corrected": 0, + "uncorrected": 0 + } + } + ], + "cooldown": { + "duration_sec": 0, + "samples": 0, + "avg_temp_c": 0, + "p95_temp_c": 0, + "avg_power_w": 0, + "p95_power_w": 0, + "avg_graphics_clock_mhz": 0, + "p95_graphics_clock_mhz": 0, + "avg_memory_clock_mhz": 0, + "p95_memory_clock_mhz": 0, + "avg_usage_pct": 0, + "avg_mem_usage_pct": 0, + "clock_cv_pct": 0, + "power_cv_pct": 0, + "temp_cv_pct": 0, + "clock_drift_pct": 0 + }, + "throttle_counters": { + "sw_power_cap_us": 100013, + "sw_thermal_slowdown_us": 100013, + "sync_boost_us": 0, + "hw_thermal_slowdown_us": 0, + "hw_power_brake_slowdown_us": 0 + }, + "ecc": { + "corrected": 0, + "uncorrected": 0 + }, + "precision_results": [ + { + "name": "fp16_tensor", + "category": "fp16_bf16", + "supported": true, + "lanes": 16, + "m": 9856, + "n": 9856, + "k": 9856, + "iterations": 28000, + "teraops_per_sec": 178.71823088298666, + "weight": 0.5, + "weighted_teraops_per_sec": 89.35911544149333 + }, + { + "name": "fp32_tf32", + "category": "fp32_tf32", + "supported": true, + "lanes": 16, + "m": 6912, + "n": 6912, + "k": 6912, + "iterations": 28000, + "teraops_per_sec": 61.64217593856, + "weight": 1, + "weighted_teraops_per_sec": 61.64217593856 + }, + { + "name": "fp64", + "category": "fp64", + "supported": false, + "weight": 2, + "notes": "unsupported" + }, + { + "name": "fp8_e4m3", + "category": "fp8", + "supported": true, + "lanes": 16, + "m": 11392, + "n": 11392, + "k": 11392, + "iterations": 28000, + "teraops_per_sec": 275.97306768042665, + "weight": 0.25, + "weighted_teraops_per_sec": 68.99326692010666 + }, + { + "name": "fp8_e5m2", + "category": "fp8", + "supported": false, + "weight": 0.25, + "notes": "unsupported" + }, + { + "name": "int8_tensor", + "category": "int8", + "supported": true, + "lanes": 16, + "m": 8832, + "n": 8832, + "k": 8832, + "iterations": 28000, + "teraops_per_sec": 128.60088385536, + "weight": 0.25, + "weighted_teraops_per_sec": 32.15022096384 + } + ], + "scores": { + "compute_score": 1269.4220752726699, + "synthetic_score": 1193.7786414934699, + "mixed_score": 252.14477926399996, + "mixed_efficiency": 0.21121568982718245, + "power_sustain_score": 0, + "thermal_sustain_score": 99.98341453537307, + "stability_score": 82.5511898186318, + "interconnect_score": 12.1, + "composite_score": 1150.5235745430427 + }, + "degradation_reasons": [ + "low_sm_clock_vs_target", + "variance_too_high" + ] + }, + { + "index": 1, + "uuid": "GPU-0182c11c-0c2c-aafd-0ada-113b64147ee6", + "name": "NVIDIA H100 80GB HBM3", + "bus_id": "00000000:C8:00.0", + "vbios": "96.00.74.00.01", + "compute_capability": "9.0", + "backend": "cublasLt", + "status": "OK", + "power_limit_w": 700, + "default_power_limit_w": 700, + "max_graphics_clock_mhz": 1980, + "max_memory_clock_mhz": 2619, + "locked_graphics_clock_mhz": 1980, + "locked_memory_clock_mhz": 2619, + "baseline": { + "duration_sec": 20.631671609, + "samples": 1, + "avg_temp_c": 69, + "p95_temp_c": 69, + "avg_power_w": 270.57, + "p95_power_w": 270.57, + "avg_graphics_clock_mhz": 1830, + "p95_graphics_clock_mhz": 1830, + "avg_memory_clock_mhz": 2619, + "p95_memory_clock_mhz": 2619, + "avg_usage_pct": 0, + "avg_mem_usage_pct": 0, + "clock_cv_pct": 0, + "power_cv_pct": 0, + "temp_cv_pct": 0, + "clock_drift_pct": 0 + }, + "steady": { + "duration_sec": 603.015967594, + "samples": 12, + "avg_temp_c": 86.16666666666667, + "p95_temp_c": 87, + "avg_power_w": 552.3258333333333, + "p95_power_w": 562.0735, + "avg_graphics_clock_mhz": 917.5, + "p95_graphics_clock_mhz": 960, + "avg_memory_clock_mhz": 2619, + "p95_memory_clock_mhz": 2619, + "avg_usage_pct": 100, + "avg_mem_usage_pct": 41.666666666666664, + "clock_cv_pct": 4.3168881519768005, + "power_cv_pct": 1.1828641347909323, + "temp_cv_pct": 0.43250831286262864, + "clock_drift_pct": 0 + }, + "precision_steady": [ + { + "precision": "int8", + "steady": { + "duration_sec": 58.789805111, + "samples": 2, + "avg_temp_c": 81.5, + "p95_temp_c": 84.65, + "avg_power_w": 618.75, + "p95_power_w": 691.767, + "avg_graphics_clock_mhz": 1185, + "p95_graphics_clock_mhz": 1239, + "avg_memory_clock_mhz": 2619, + "p95_memory_clock_mhz": 2619, + "avg_usage_pct": 100, + "avg_mem_usage_pct": 79, + "clock_cv_pct": 5.063291139240507, + "power_cv_pct": 13.111919191919192, + "temp_cv_pct": 4.294478527607362, + "clock_drift_pct": 0 + }, + "teraops_per_sec": 994.2685477502977, + "weighted_teraops_per_sec": 248.56713693757442, + "ecc": { + "corrected": 0, + "uncorrected": 0 + } + }, + { + "precision": "fp8", + "steady": { + "duration_sec": 107.153140143, + "samples": 2, + "avg_temp_c": 86.5, + "p95_temp_c": 86.95, + "avg_power_w": 595.51, + "p95_power_w": 611.8629999999999, + "avg_graphics_clock_mhz": 960, + "p95_graphics_clock_mhz": 973.5, + "avg_memory_clock_mhz": 2619, + "p95_memory_clock_mhz": 2619, + "avg_usage_pct": 100, + "avg_mem_usage_pct": 53, + "clock_cv_pct": 1.5625, + "power_cv_pct": 3.0511662272673776, + "temp_cv_pct": 0.5780346820809248, + "clock_drift_pct": 0 + }, + "teraops_per_sec": 1073.9294805164031, + "weighted_teraops_per_sec": 268.4823701291008, + "ecc": { + "corrected": 0, + "uncorrected": 0 + } + }, + { + "precision": "fp16", + "steady": { + "duration_sec": 162.145217446, + "samples": 3, + "avg_temp_c": 87, + "p95_temp_c": 87.9, + "avg_power_w": 554.0466666666666, + "p95_power_w": 604.477, + "avg_graphics_clock_mhz": 945, + "p95_graphics_clock_mhz": 999, + "avg_memory_clock_mhz": 2619, + "p95_memory_clock_mhz": 2619, + "avg_usage_pct": 100, + "avg_mem_usage_pct": 34.666666666666664, + "clock_cv_pct": 5.184105275731594, + "power_cv_pct": 10.324119223592762, + "temp_cv_pct": 0.9385018171583058, + "clock_drift_pct": 0 + }, + "teraops_per_sec": 560.1539979389611, + "weighted_teraops_per_sec": 280.07699896948054, + "ecc": { + "corrected": 0, + "uncorrected": 0 + } + }, + { + "precision": "fp32", + "steady": { + "duration_sec": 221.67717378, + "samples": 3, + "avg_temp_c": 86, + "p95_temp_c": 86, + "avg_power_w": 565.8233333333334, + "p95_power_w": 594, + "avg_graphics_clock_mhz": 1075, + "p95_graphics_clock_mhz": 1092, + "avg_memory_clock_mhz": 2619, + "p95_memory_clock_mhz": 2619, + "avg_usage_pct": 100, + "avg_mem_usage_pct": 35, + "clock_cv_pct": 1.315547499881949, + "power_cv_pct": 4.1295451621968935, + "temp_cv_pct": 0, + "clock_drift_pct": 0 + }, + "teraops_per_sec": 293.76899847290883, + "weighted_teraops_per_sec": 293.76899847290883, + "ecc": { + "corrected": 0, + "uncorrected": 0 + } + }, + { + "precision": "fp64", + "steady": { + "duration_sec": 283.877134755, + "samples": 3, + "avg_temp_c": 87, + "p95_temp_c": 87, + "avg_power_w": 565.93, + "p95_power_w": 581.971, + "avg_graphics_clock_mhz": 1055, + "p95_graphics_clock_mhz": 1077, + "avg_memory_clock_mhz": 2619, + "p95_memory_clock_mhz": 2619, + "avg_usage_pct": 100, + "avg_mem_usage_pct": 42, + "clock_cv_pct": 1.7732973397032898, + "power_cv_pct": 2.3474912323387347, + "temp_cv_pct": 0, + "clock_drift_pct": 0 + }, + "ecc": { + "corrected": 0, + "uncorrected": 0 + } + }, + { + "precision": "fp4", + "steady": { + "duration_sec": 353.430169396, + "samples": 3, + "avg_temp_c": 87, + "p95_temp_c": 87, + "avg_power_w": 547.1333333333333, + "p95_power_w": 548.109, + "avg_graphics_clock_mhz": 885, + "p95_graphics_clock_mhz": 939, + "avg_memory_clock_mhz": 2619, + "p95_memory_clock_mhz": 2619, + "avg_usage_pct": 100, + "avg_mem_usage_pct": 42.333333333333336, + "clock_cv_pct": 5.535570040187973, + "power_cv_pct": 0.15226269553094146, + "temp_cv_pct": 0, + "clock_drift_pct": 0 + }, + "ecc": { + "corrected": 0, + "uncorrected": 0 + } + } + ], + "cooldown": { + "duration_sec": 0, + "samples": 0, + "avg_temp_c": 0, + "p95_temp_c": 0, + "avg_power_w": 0, + "p95_power_w": 0, + "avg_graphics_clock_mhz": 0, + "p95_graphics_clock_mhz": 0, + "avg_memory_clock_mhz": 0, + "p95_memory_clock_mhz": 0, + "avg_usage_pct": 0, + "avg_mem_usage_pct": 0, + "clock_cv_pct": 0, + "power_cv_pct": 0, + "temp_cv_pct": 0, + "clock_drift_pct": 0 + }, + "throttle_counters": { + "sw_power_cap_us": 0, + "sw_thermal_slowdown_us": 0, + "sync_boost_us": 0, + "hw_thermal_slowdown_us": 0, + "hw_power_brake_slowdown_us": 0 + }, + "ecc": { + "corrected": 0, + "uncorrected": 0 + }, + "precision_results": [ + { + "name": "fp16_tensor", + "category": "fp16_bf16", + "supported": true, + "lanes": 16, + "m": 9856, + "n": 9856, + "k": 9856, + "iterations": 25760, + "teraops_per_sec": 164.42077241234773, + "weight": 0.5, + "weighted_teraops_per_sec": 82.21038620617387 + }, + { + "name": "fp32_tf32", + "category": "fp32_tf32", + "supported": true, + "lanes": 16, + "m": 6912, + "n": 6912, + "k": 6912, + "iterations": 25760, + "teraops_per_sec": 56.710801863475204, + "weight": 1, + "weighted_teraops_per_sec": 56.710801863475204 + }, + { + "name": "fp64", + "category": "fp64", + "supported": false, + "weight": 2, + "notes": "unsupported" + }, + { + "name": "fp8_e4m3", + "category": "fp8", + "supported": true, + "lanes": 16, + "m": 11392, + "n": 11392, + "k": 11392, + "iterations": 25760, + "teraops_per_sec": 253.89522226599254, + "weight": 0.25, + "weighted_teraops_per_sec": 63.473805566498136 + }, + { + "name": "fp8_e5m2", + "category": "fp8", + "supported": false, + "weight": 0.25, + "notes": "unsupported" + }, + { + "name": "int8_tensor", + "category": "int8", + "supported": true, + "lanes": 16, + "m": 8832, + "n": 8832, + "k": 8832, + "iterations": 25760, + "teraops_per_sec": 118.3128131469312, + "weight": 0.25, + "weighted_teraops_per_sec": 29.5782032867328 + } + ], + "scores": { + "compute_score": 1160.4874635859287, + "synthetic_score": 1090.8955045090647, + "mixed_score": 231.97319692288, + "mixed_efficiency": 0.21264474549950116, + "power_sustain_score": 0, + "thermal_sustain_score": 100, + "stability_score": 75.33829122588782, + "interconnect_score": 12.1, + "composite_score": 1030.9140807492215 + }, + "degradation_reasons": [ + "low_sm_clock_vs_target", + "variance_too_high" + ] + } + ], + "interconnect": { + "status": "OK", + "attempted": true, + "supported": true, + "selected_gpu_indices": [ + 0, + 1 + ], + "avg_algbw_gbps": 11.920000000000002, + "max_algbw_gbps": 12.1, + "avg_busbw_gbps": 11.920000000000002, + "max_busbw_gbps": 12.1 + }, + "server_power": { + "available": false, + "notes": [ + "IPMI power reading unavailable; server-side power characterization skipped" + ] + } +} \ No newline at end of file