Cisco · HPF HGX System (32xH100-SXM-80GB, TensorRT)
Llama-3.1-405B (MLPerf reference) · llama3.1-405b · Server · run 01a00b49-29b8…
valid · Reported evidence
Measurements
output_token_throughput_tps · reported1,126.94 tokens_s
Llama-3.1-405B (MLPerf reference) · llama3.1-405b · Server · server · 8× NVIDIA H100-SXM-80GB × 4 nodes · mlperf_closed_99 · observed 2025-09-04
Identity
modelLlama-3.1-405B (MLPerf reference)
stackTensorRT 10.8, CUDA 12.8
workloadllama3.1-405b · Server
topology8× NVIDIA H100-SXM-80GB × 4 nodes
harnessmlperf_loadgen v5.1
quality policymlperf_closed_99
comparison keysha256:57933320d943b947…
sha256:e26bc4b373072de44cfd686…
Launch configuration
dtypefp8
systemHPF HGX System (32xH100-SXM-80GB, TensorRT)
softwareTensorRT 10.8, CUDA 12.8
availabilityavailable
Workload
namellama3.1-405b · Server
streamingyes
load generationopen_loop
Canonical manifest
Show manifestHide manifest
{
"kind": "ServingRun",
"spec": {
"status": "valid",
"harness": {
"name": "mlperf_loadgen",
"version": "v5.1"
},
"topology": {
"nodeCount": 4,
"acceleratorModel": "NVIDIA H100-SXM-80GB",
"acceleratorVendor": "nvidia",
"acceleratorsPerNode": 8
},
"observedAt": "2025-09-04T19:30:14Z",
"modelDigest": "sha256:bc69ffe83643f1c91a137a177c5ef85829a1f5b65ae51771adb75ef783baf031",
"measurements": [
{
"unit": "tokens_s",
"value": 1126.94,
"metric": "output_token_throughput_tps",
"statistic": "reported"
}
],
"sourceNative": {
"source": "mlperf-inference",
"metrics": {
"performance_units": "Tokens/s",
"performance_result": 1126.94
},
"category": "datacenter",
"division": "closed",
"scenario": "server",
"benchmark": "llama3.1-405b",
"externalId": "v5.1/5.1-0016"
},
"qualityPolicy": "mlperf_closed_99",
"workloadDigest": "sha256:3428d964054278646d17c2f6064b4e29982b970b7fa607362cf50b7f7dff994c",
"configurationDigest": "sha256:9335e6a1ab65bbc3a17837505d8091248c6d37726e21320ff66284f962bb54e1"
},
"metadata": {
"name": "mlperf-v5-1-5-1-0016",
"title": "Cisco · llama3.1-405b · Server",
"sourceRefs": [
{
"url": "https://github.com/mlcommons/inference_results_v5.1/tree/main/closed/Cisco/results/HGX-H100_H100-SXM-80GBx32_TRT"
}
]
},
"apiVersion": "kernelindex.dev/v1alpha1"
}Cite this result (permalink, digest, access date)
Report an issue with this serving run
MLPerf Inference · closed/datacenter · entry v5.1/5.1-0016 · shown unmodified as publishedAll serving results →JSON