2026-04-13 benchmarks
This commit is contained in:
@@ -0,0 +1,6 @@
|
||||
# Allow only result.json and README.md everywhere; block everything else.
|
||||
*
|
||||
!*/
|
||||
!.gitignore
|
||||
!README.md
|
||||
!**/result.json
|
||||
@@ -0,0 +1,127 @@
|
||||
# pub-beebench
|
||||
|
||||
Публичный репозиторий с результатами GPU-бенчмарков. Содержимое автоматически
|
||||
публикуется на сайте через [beebench-publisher](https://github.com/reanimator/beebench).
|
||||
|
||||
---
|
||||
|
||||
## Структура репозитория
|
||||
|
||||
Каждый запуск бенчмарка — отдельная **директория** в корне репозитория или
|
||||
архив `.tar.gz`. Вложенность произвольная: `result.json` может лежать как
|
||||
непосредственно в директории бандла, так и в поддиректории.
|
||||
|
||||
```
|
||||
pub-beebench/
|
||||
dell-r760-h100-2026-04-12/
|
||||
result.json
|
||||
00-nvidia-smi-q.log # опционально — используется для определения
|
||||
gpu-0-steady.log # модели GPU если поле name пустое в result.json
|
||||
supermicro-x13-a100-2026-03-01/
|
||||
logs/
|
||||
gpu-0-steady.log
|
||||
result.json
|
||||
xe9680-h100x8-2026-02-15.tar.gz
|
||||
```
|
||||
|
||||
**Правила именования директории/архива** — произвольные, на публикацию не
|
||||
влияют. Рекомендуется использовать читаемые имена с датой:
|
||||
`<сервер>-<gpu>-<дата>`.
|
||||
|
||||
---
|
||||
|
||||
## Формат result.json
|
||||
|
||||
Обязательные поля отмечены `*`.
|
||||
|
||||
```jsonc
|
||||
{
|
||||
"benchmark_version": "v1.0.0",
|
||||
"generated_at": "2026-04-12T10:30:00Z", // * ISO 8601 UTC
|
||||
"server_model": "Dell PowerEdge R760", // * используется как заголовок страницы сервера
|
||||
"hostname": "r760-lab-01", // отображается если server_model не задан
|
||||
"benchmark_profile": "standard", // произвольная строка: standard / overnight / …
|
||||
"overall_status": "OK", // OK | WARN | FAIL
|
||||
"selected_gpu_indices": [0], // индексы GPU участвовавших в тесте
|
||||
"findings": ["Stable benchmark run"], // список наблюдений
|
||||
"warnings": [],
|
||||
"gpus": [ // * минимум один элемент
|
||||
{
|
||||
"index": 0, // * порядковый номер GPU
|
||||
"name": "NVIDIA H100 PCIe 80GB", // * модель; если пустая — берётся из лог-файлов
|
||||
"status": "OK", // OK | WARN | FAIL
|
||||
"bus_id": "0000:17:00.0",
|
||||
"vbios": "96.00.5E.00.02",
|
||||
"compute_capability": "9.0",
|
||||
"multiprocessor_count": 114,
|
||||
"steady": {
|
||||
"duration_sec": 90,
|
||||
"avg_temp_c": 73.2,
|
||||
"p95_temp_c": 75.0,
|
||||
"avg_power_w": 311.5,
|
||||
"p95_power_w": 321.0,
|
||||
"avg_graphics_clock_mhz": 1618,
|
||||
"p95_graphics_clock_mhz": 1634,
|
||||
"avg_memory_clock_mhz": 1593,
|
||||
"p95_memory_clock_mhz": 1593,
|
||||
"avg_usage_pct": 98.4,
|
||||
"avg_mem_usage_pct": 92.1
|
||||
},
|
||||
"scores": { // * все поля обязательны; compute_score > 0
|
||||
"compute_score": 932.4,
|
||||
"power_sustain_score": 901.2,
|
||||
"thermal_sustain_score": 889.3,
|
||||
"stability_score": 944.0,
|
||||
"composite_score": 926.7
|
||||
},
|
||||
"degradation_reasons": [],
|
||||
"notes": ["Healthy run"]
|
||||
}
|
||||
]
|
||||
}
|
||||
```
|
||||
|
||||
### Обязательные условия для публикации
|
||||
|
||||
- `generated_at` — должно быть заполнено и являться валидной датой ISO 8601
|
||||
- `gpus` — минимум один элемент
|
||||
- `compute_score` хотя бы у одного GPU должен быть больше `0`; запуски где все
|
||||
оценки равны нулю считаются неудавшимися и отклоняются
|
||||
|
||||
---
|
||||
|
||||
## Несколько GPU в одном запуске
|
||||
|
||||
Если сервер тестировался с несколькими GPU одновременно, все они перечисляются
|
||||
в массиве `gpus`. Каждая карта получает отдельную запись с собственными
|
||||
метриками и оценками.
|
||||
|
||||
```jsonc
|
||||
"gpus": [
|
||||
{ "index": 0, "name": "NVIDIA H100 PCIe 80GB", ... },
|
||||
{ "index": 1, "name": "NVIDIA H100 PCIe 80GB", ... }
|
||||
]
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## Дополнительные файлы в бандле
|
||||
|
||||
Если поле `name` у GPU пустое, publisher попытается определить модель из
|
||||
следующих файлов (ищутся в той же директории что и `result.json`):
|
||||
|
||||
| Файл | Источник |
|
||||
|---|---|
|
||||
| `00-nvidia-smi-q.log` | вывод `nvidia-smi -q` |
|
||||
| `gpu-0-warmup.log` | лог прогрева |
|
||||
| `gpu-0-steady.log` | лог steady-фазы |
|
||||
|
||||
Остальные файлы (`.html`, `.csv`, `.txt`) игнорируются но могут присутствовать.
|
||||
|
||||
---
|
||||
|
||||
## Удаление результата
|
||||
|
||||
Чтобы убрать результат с сайта — удалите соответствующую директорию или архив
|
||||
из репозитория и сделайте коммит. Сайт обновится при следующей синхронизации
|
||||
(по умолчанию раз в сутки).
|
||||
@@ -0,0 +1,190 @@
|
||||
{
|
||||
"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
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,339 @@
|
||||
{
|
||||
"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
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,485 @@
|
||||
{
|
||||
"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
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,631 @@
|
||||
{
|
||||
"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
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,776 @@
|
||||
{
|
||||
"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
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,922 @@
|
||||
{
|
||||
"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
|
||||
}
|
||||
}
|
||||
File diff suppressed because it is too large
Load Diff
File diff suppressed because it is too large
Load Diff
Reference in New Issue
Block a user