Llama 4 Scout on 8×H100 SXM with vLLM (public benchmark)

Submitted by @evokernel-bot on 2026-04-28 · https://evokernel.dev/en/cases/case-llama4-scout-h100x8-vllm-001/

Stack

Hardware
h100-sxm5 × 8 (single-node-hgx)
Server
nvidia-hgx-h100
Interconnect
intra: nvlink-4 · inter: none
Model
llama-4-scout (bf16)
Engine
vllm0.6.0
Quantization
bf16
Parallel
TP=8 · PP=1 · EP=1 · SP=1
Driver
CUDA 12.4
OS
Ubuntu 22.04

Scenario

Prefill seq
1024
Decode seq
256
Batch
16
Max concurrent
64

Results

Decode tok/s
1850
Prefill tok/s
26000
TTFT p50
ms
145
TBT p50
ms
18
Memory/card
GB
28
Power/card
W
580
Compute
util %
48
Memory BW
util %
62

Same-model side-by-side

本 case vs 同模型其他 case 的吞吐对比

Bottleneck — memory-bandwidth

Compute 48% Memory BW 62% Other 0%

Reproduction

vllm serve meta-llama/Llama-4-Scout --tensor-parallel-size 8 --max-model-len 16384

Benchmark tool: vllm benchmark_serving.py + sharegpt

Optimization patterns

memory-bound-decode-prefer-int8

Citations

  1. [1] vLLM official Llama 4 Scout benchmark notes (figures approximate from blog) — https://blog.vllm.ai/ · 2026-04-28 实测验证
    Attestation: Numbers extracted from public vLLM benchmark blog; not independently re-run by submitter.