inference/research
verifiedmedium · 0.75updated 2026-10-15

Argonne MI300X vLLM Llama-3-70B b1

51.4 tok/s aggregate (batch*(in+out)/lat)

Fields

id
benchmarks/argonne-mi300x-vllm-llama3-70b-fp16-b1-128io-tps
Metric
tps_aggregate
Value
51.4
Unit
tok/s aggregate (batch*(in+out)/lat)
Model
meta-llama/Meta-Llama-3-70B
Methodology
THIRD-PARTY MEASURED (Argonne LLM-Inference-Bench arXiv:2411.00136, All_results.csv). vLLM, FP16/BF16, 128in/128out, AGGREGATE over batch. 1x bs1 lat 4.98s. CSV 'AMD MI300X GPU' = 8x MI300X node (xGMI Infinity Fabric). 70B FP16 fits in one 192GB MI300X.
Measured by
third_party
Reproducible
yes

Cross-references