58 lines
1.7 KiB
JSON
58 lines
1.7 KiB
JSON
{
|
|
"benchmark": "t2va-dialogue-quoted-864x480-141f-base12-sage2-seed440420.json",
|
|
"actual_attention": "sdpa",
|
|
"mode": "tensor",
|
|
"world_size": 1,
|
|
"resolution": [
|
|
864,
|
|
480
|
|
],
|
|
"frames": 141,
|
|
"steps": 12,
|
|
"seed": 440420,
|
|
"rtx_pro_6000_blackwell_server": {
|
|
"cloud": "RunPod Secure",
|
|
"data_center": "EUR-IS-1",
|
|
"gpu_memory_mib": 97887,
|
|
"driver": "595.91.07",
|
|
"host_cuda": "13.2",
|
|
"torch": "2.9.1+cu130",
|
|
"hourly_usd": 2.09,
|
|
"sampling_seconds": [
|
|
28.431582752993563,
|
|
28.57024686699151
|
|
],
|
|
"sampling_mean_seconds": 28.500914809992537,
|
|
"sampling_range_percent": 0.48652513409615622,
|
|
"model_load_seconds": [
|
|
4.332073616009438,
|
|
4.819102452020161
|
|
],
|
|
"conditioning_seconds": [
|
|
8.492386644007638,
|
|
8.663792312989244
|
|
],
|
|
"peak_sampling_allocated_bytes": 14049528832,
|
|
"checksums": [
|
|
98329.34375,
|
|
-345.9080505371094
|
|
],
|
|
"persistent_reports": [
|
|
"/runpod-volume/h3-benchmarks/rtxpro6000-server-1gpu-sdpa-864x480-141f-base12-seed440420.json",
|
|
"/runpod-volume/h3-benchmarks/rtxpro6000-server-1gpu-sdpa-864x480-141f-base12-seed440420-repeat2.json"
|
|
]
|
|
},
|
|
"gb10": {
|
|
"direct_sdpa_sampling_seconds": 125.3,
|
|
"same_tensor_runner_sampling_seconds": 126.66326454500086
|
|
},
|
|
"speedup": {
|
|
"versus_gb10_direct_sdpa": 4.396350111403066,
|
|
"versus_gb10_same_tensor_runner": 4.444182419737356
|
|
},
|
|
"notes": [
|
|
"Both RTX PRO 6000 runs produced identical checksums.",
|
|
"No latent was retained, so this is a performance and execution-stability gate rather than a cross-device numerical parity gate.",
|
|
"The billable pod was terminated after the repeat run; the network volume and reports were preserved."
|
|
]
|
|
}
|