MLPerf v6.0 HPE 5x NVIDIA RTX PRO 4500 Blackwell 32GB: llama3.1-8b Offline aggregate tok/s
11665.0 output tokens/s (aggregate over 5 accelerators, Offline)
Fields
id
benchmarks/mlperf-v6-0-hpe-5xrtxpro4500-llama31-8b-llama3-1-8b-offlineMetric
tps_aggregate
Value
11,665
Unit
output tokens/s (aggregate over 5 accelerators, Offline)
Model
llama3.1-8b
Methodology
VENDOR-SUBMITTED MLPerf Inference v6.0 closed division, Datacenter suite, Offline scenario. HPE ProLiant DL380 5xRTX PRO 4500 PCIe-32GB: 5x NVIDIA RTX PRO 4500 Blackwell 32GB. Framework TensorRT 10.11.0.33 CUDA 13.0. 'Tokens/s' = AGGREGATE output-token throughput across the 5 accelerators (NOT per-request). Value verbatim from summary_results.json. min_query_count/min_duration per MLPerf rules; binding latency constraint applies to Server/Interactive.
Measured by
vendor
Reproducible
yes