Speed test matrix

Qwen3.6-27B-AWQ on CMP 170HX 64GB

vllm · 0.32.0rc1.dev249+g18305948a · AWQ W4 g1286 contexts (0–248K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 128K

by @Sandbergllm-inference-bench v0.8.0Measured Nov 1, 2026, 11:23 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: AWQ W4 g128
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 QuantTrio/Qwen3.6-27B-AWQ --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 context1,9594,157 ms TTFT

16K context1,8538,698 ms TTFT

32K context1,73618,486 ms TTFT

64K context1,53341,769 ms TTFT

128K context1,242102,941 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	50.8156 ms TTFT	99.0271 ms TTFT	189.8328 ms TTFT	354.4532 ms TTFT	583.8858 ms TTFT	906.31,456 ms TTFT	1,161.12,730 ms TTFT	Over capacity
16K	48.3396 ms TTFT	90.4562 ms TTFT	162.8885 ms TTFT	275.01,523 ms TTFT	400.92,815 ms TTFT	Over capacity	Over capacity	Over capacity
32K	46.2673 ms TTFT	83.2940 ms TTFT	142.61,541 ms TTFT	224.62,656 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
64K	42.4765 ms TTFT	71.61,072 ms TTFT	114.31,661 ms TTFT	164.52,821 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
128K	36.3963 ms TTFT	56.11,284 ms TTFT	81.81,765 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
248K	26.8Requests queued2,614 ms TTFT	35.6Requests queued3,387 ms TTFT	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	1,959.0	—	4,157 ms	—	—	1	—	—
Prefill	16K	1	Valid	1,853.0	—	8,698 ms	—	—	1	—	—
Prefill	32K	1	Valid	1,736.0	—	18,486 ms	—	—	1	—	—
Prefill	64K	1	Valid	1,533.0	—	41,769 ms	—	—	1	—	—
Prefill	128K	1	Valid	1,242.0	—	102,941 ms	—	—	1	—	—
Sustained decode	0	1	Valid	50.8	50.8	156 ms	19.7 ms	1.0	0	1,523	30.0 s
Sustained decode	0	2	Valid	99.0	49.6	271 ms	20.2 ms	2.0	0	2,968	30.0 s
Sustained decode	0	4	Valid	189.8	47.5	328 ms	21.0 ms	4.0	0	5,692	30.0 s
Sustained decode	0	8	Valid	354.4	44.4	532 ms	22.5 ms	8.0	0	10,632	30.0 s
Sustained decode	0	16	Valid	583.8	36.6	858 ms	27.3 ms	16.0	0	17,504	30.0 s
Sustained decode	0	32	Valid	906.3	28.4	1,456 ms	35.2 ms	32.0	0	27,168	30.0 s
Sustained decode	0	64	Valid	1,161.1	18.1	2,730 ms	55.3 ms	64.0	0	34,816	30.0 s
Sustained decode	0	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	48.3	48.4	396 ms	20.7 ms	1.0	0	1,450	30.0 s
Sustained decode	16K	2	Valid	90.4	45.1	562 ms	22.2 ms	2.0	0	2,710	30.0 s
Sustained decode	16K	4	Valid	162.8	40.3	885 ms	24.8 ms	4.0	0	4,884	30.0 s
Sustained decode	16K	8	Valid	275.0	33.5	1,523 ms	29.8 ms	8.0	0	8,248	30.0 s
Sustained decode	16K	16	Valid	400.9	23.6	2,815 ms	42.4 ms	16.0	0	12,016	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	46.2	46.2	673 ms	21.6 ms	1.0	0	1,385	30.0 s
Sustained decode	32K	2	Valid	83.2	41.4	940 ms	24.2 ms	2.0	0	2,496	30.0 s
Sustained decode	32K	4	Valid	142.6	35.0	1,541 ms	28.6 ms	4.0	0	4,276	30.0 s
Sustained decode	32K	8	Valid	224.6	26.8	2,656 ms	37.4 ms	8.0	0	6,736	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	42.4	42.5	765 ms	23.5 ms	1.0	0	1,271	30.0 s
Sustained decode	64K	2	Valid	71.6	35.7	1,072 ms	28.0 ms	2.0	0	2,148	30.0 s
Sustained decode	64K	4	Valid	114.3	28.1	1,661 ms	35.6 ms	4.0	0	3,428	30.0 s
Sustained decode	64K	8	Valid	164.5	19.6	2,821 ms	51.0 ms	8.0	0	4,930	30.0 s
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	36.3	36.5	963 ms	27.4 ms	1.0	0	1,090	30.0 s
Sustained decode	128K	2	Valid	56.1	28.1	1,284 ms	35.6 ms	2.0	0	1,682	30.0 s
Sustained decode	128K	4	Valid	81.8	20.3	1,765 ms	49.2 ms	4.0	0	2,452	30.0 s
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	26.8	28.8	2,614 ms	34.7 ms	0.9	0	804	30.0 s
Sustained decode	248K	2	Requests queued	35.6	19.7	3,387 ms	50.8 ms	1.8	0	1,068	30.0 s
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.