Speed test matrix

Qwen3.8-27B-INT8-W8A16-MTP on CMP 170HX 64GB

vllm · 0.32.0rc1.dev249+g18305948a · INT8 W8A166 contexts (0–248K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 128K

by @Sandbergllm-inference-bench v0.8.0Measured Oct 30, 2026, 16:49 UTC

Hardware

Type: Discrete GPU
GPU: CMP 170HX
VRAM: 64 GB
CPU: 2x Intel Xeon E5-2630 v3
RAM: 64 GB
OS: Ubuntu 26.04.1

Engine

Engine: vllm
Version: 0.32.0rc1.dev249+g18305948a
Quantization: INT8 W8A16
Backend: cuda
Repository: github.com/vllm-project/vllm
Build: ghcr.io/morrowmake/vllm-cmp170hx@sha256:5b45c4d6feafefe2eba962375b0a9a0b64fd22b9e05abb4d1966f14a32a2db79

Run

Served model: bench
Revision: main
Benchmark tool: llm-inference-bench v0.8.0
Decode test: Fixed duration
Prefill mode: Measured during decode warm-up
Time per measurement: 30 s
Max output: 8,192 tokens

Launch command

# Untuned out-of-the-box baseline. Spec decode: none
# 1x NVIDIA CMP 170HX 64GB (cmpunlocker), 180 W limit, PCIe Gen2 x4 | 2x Intel Xeon E5-2630 v3 | 64 GB DDR4 | Ubuntu 26.04.1, NVIDIA driver 610.57.04 | vLLM: Morrowmake vllm-cmp170hx image
vllm serve lued/Qwen3.8-27B-INT8-W8A16-MTP --max-model-len 262144 --gpu-memory-utilization 0.94 --max-num-seqs 128 --enable-prefix-caching --trust-remote-code --language-model-only

Prefill

prompt tok/s by context length

8K context1,6644,923 ms TTFT

16K context1,56810,310 ms TTFT

32K context1,49121,547 ms TTFT

64K context1,33847,882 ms TTFT

128K context1,107115,549 ms TTFT

Sustained decode

total tok/s · rows are context length, columns are simultaneous requests

Context	1	2	4	8	16	32	64	128
0	37.0181 ms TTFT	72.3228 ms TTFT	139.6397 ms TTFT	266.1713 ms TTFT	477.41,321 ms TTFT	766.42,107 ms TTFT	Over capacity	Over capacity
16K	35.5455 ms TTFT	67.8675 ms TTFT	124.81,068 ms TTFT	218.11,872 ms TTFT	343.23,507 ms TTFT	Over capacity	Over capacity	Over capacity
32K	34.4762 ms TTFT	63.91,101 ms TTFT	112.81,746 ms TTFT	184.03,068 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
64K	32.4882 ms TTFT	57.31,222 ms TTFT	94.41,845 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
128K	29.01,122 ms TTFT	47.21,333 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
248K	22.3Requests queued2,838 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity

Requests queued: the server could not run all requests at once, so some waited in a queue. Over capacity: not run because this many requests at this context would exceed the server's KV cache.

TTFT is measured after the benchmark warms the server's prompt cache, so it mostly excludes prefill.

All measurements

Latencies are medians. Total tok/s is output across all simultaneous requests.

Phase	Context	Requests at once	Status	Total tok/s	Per request tok/s	TTFT	Time between tokens	Average active requests	Completed requests	Output tokens	Measured for
Prefill	8K	1	Valid	1,664.0	—	4,923 ms	—	—	1	—	—
Prefill	16K	1	Valid	1,568.0	—	10,310 ms	—	—	1	—	—
Prefill	32K	1	Valid	1,491.0	—	21,547 ms	—	—	1	—	—
Prefill	64K	1	Valid	1,338.0	—	47,882 ms	—	—	1	—	—
Prefill	128K	1	Valid	1,107.0	—	115,549 ms	—	—	1	—	—
Sustained decode	0	1	Valid	37.0	37.0	181 ms	27.0 ms	1.0	0	1,108	30.0 s
Sustained decode	0	2	Valid	72.3	36.2	228 ms	27.6 ms	2.0	0	2,166	30.0 s
Sustained decode	0	4	Valid	139.6	34.9	397 ms	28.6 ms	4.0	0	4,184	30.0 s
Sustained decode	0	8	Valid	266.1	33.3	713 ms	30.0 ms	8.0	0	7,976	30.0 s
Sustained decode	0	16	Valid	477.4	29.9	1,321 ms	33.5 ms	16.0	0	14,320	30.0 s
Sustained decode	0	32	Valid	766.4	23.7	2,107 ms	42.2 ms	32.0	0	23,008	30.0 s
Sustained decode	0	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	35.5	35.6	455 ms	28.1 ms	1.0	0	1,065	30.0 s
Sustained decode	16K	2	Valid	67.8	33.7	675 ms	29.6 ms	2.0	0	2,032	30.0 s
Sustained decode	16K	4	Valid	124.8	30.8	1,068 ms	32.4 ms	4.0	0	3,744	30.0 s
Sustained decode	16K	8	Valid	218.1	26.4	1,872 ms	37.9 ms	8.0	0	6,536	30.0 s
Sustained decode	16K	16	Valid	343.2	19.9	3,507 ms	50.1 ms	16.0	0	10,288	30.0 s
Sustained decode	16K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	32K	1	Valid	34.4	34.5	762 ms	29.0 ms	1.0	0	1,033	30.0 s
Sustained decode	32K	2	Valid	63.9	31.7	1,101 ms	31.5 ms	2.0	0	1,916	30.0 s
Sustained decode	32K	4	Valid	112.8	27.5	1,746 ms	36.3 ms	4.0	0	3,380	30.0 s
Sustained decode	32K	8	Valid	184.0	21.7	3,068 ms	46.0 ms	8.0	0	5,521	30.0 s
Sustained decode	32K	16	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	32K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	32K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	32K	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	64K	1	Valid	32.4	32.5	882 ms	30.8 ms	1.0	0	972	30.0 s
Sustained decode	64K	2	Valid	57.3	28.5	1,222 ms	35.1 ms	2.0	0	1,716	30.0 s
Sustained decode	64K	4	Valid	94.4	23.1	1,845 ms	43.3 ms	4.0	0	2,832	30.0 s
Sustained decode	64K	8	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	64K	16	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	64K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	64K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	64K	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	128K	1	Valid	29.0	29.1	1,122 ms	34.4 ms	1.0	0	869	30.0 s
Sustained decode	128K	2	Valid	47.2	23.6	1,333 ms	42.4 ms	2.0	0	1,416	30.0 s
Sustained decode	128K	4	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	128K	8	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	128K	16	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	128K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	128K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	128K	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	248K	1	Requests queued	22.3	24.2	2,838 ms	41.4 ms	0.9	0	669	30.0 s
Sustained decode	248K	2	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	248K	4	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	248K	8	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	248K	16	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	248K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	248K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	248K	128	Over capacity	—	—	—	—	—	—	—	—

Requests queued: the server could not run all requests at once, so some waited in a queue. Over capacity: not run because this many requests at this context would exceed the server's KV cache.