{ "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" ] } }