Argonne: A100 40GB SXM x4 TP, DeepSpeed-MII, Llama-2-70B FP16, 128io, b64
1804.2 tok/s aggregate (batch*(in+out)/latency)
Fields
id
benchmarks/argonne-a100sxm40-dsmii-llama2-70b-fp16-4x-b64-128io-tpsMetric
tps_aggregate
Value
1,804.2
Unit
tok/s aggregate (batch*(in+out)/latency)
Model
meta-llama/Llama-2-70b-hf
Methodology
THIRD-PARTY MEASURED (Argonne LLM-Inference-Bench, arXiv:2411.00136, All_results.csv). DeepSpeed-MII, FP16/BF16, 128 in / 128 out. Throughput = batch*(in+out)/end-to-end latency (paper Eq.2) = AGGREGATE over the batch, NOT per-request decode. Batch 64, 4x A100-40GB-SXM4 TP. Latency 9.081s.
Measured by
third_party
Reproducible
yes
notes
DS-MII ~1.4x behind TRT-LLM at bs64. CSV column 'Nvidia A100 GPU' = ALCF ThetaGPU node (4x A100 40GB SXM4 + NVSwitch, paper Table II). Filed under nvidia-a100-40gb-sxm4 ONLY; the PCIe 40GB SKU shares the 1555 GB/s HBM2 but was not the measured part, so nvidia-a100-40gb-pcie stays uncovered.
Notes
DS-MII ~1.4x behind TRT-LLM at bs64. CSV column 'Nvidia A100 GPU' = ALCF ThetaGPU node (4x A100 40GB SXM4 + NVSwitch, paper Table II). Filed under nvidia-a100-40gb-sxm4 ONLY; the PCIe 40GB SKU shares the 1555 GB/s HBM2 but was not the measured part, so nvidia-a100-40gb-pcie stays uncovered.