Argonne LLM-Inference-Bench: MI250 x1, vLLM 0.6.1, Llama-3-8B FP16, 128in/128out, batch 1
130.2 tok/s aggregate (batch 1, single request)
Fields
id
benchmarks/argonne-mi250-vllm-llama3-8b-fp16-b1-128io-tpsMetric
tps_aggregate
Value
130.2
Unit
tok/s aggregate (batch 1, single request)
Model
meta-llama/Meta-Llama-3-8B
Methodology
THIRD-PARTY MEASURED (Argonne LLM-Inference-Bench, arXiv:2411.00136, All_results.csv). vLLM benchmark_throughput.py, FP16/BF16, 128 in / 128 out. Throughput = batch*(in+out)/end-to-end latency = AGGREGATE over the batch, NOT per-request decode. Batch 1, 1x MI250. Latency 1.967s.
Measured by
third_party
Reproducible
yes
notes
GQA model slightly slower than Llama-2-7B on MI250 at batch 1, consistent with the paper's note that MI250 saturates earlier than A100 under NUMA/page-fault handling.
Notes
GQA model slightly slower than Llama-2-7B on MI250 at batch 1, consistent with the paper's note that MI250 saturates earlier than A100 under NUMA/page-fault handling.