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, 12:04 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: MTP, 2 tokens lookahead
# 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 --speculative-config {"method":"mtp","num_speculative_tokens":2}

Prefill

prompt tok/s by context length

8K context1,9214,239 ms TTFT

16K context1,8218,848 ms TTFT

32K context1,71318,730 ms TTFT

64K context1,50942,419 ms TTFT

128K context1,219104,896 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	85.1167 ms TTFT	162.5226 ms TTFT	274.6379 ms TTFT	451.4602 ms TTFT	689.8941 ms TTFT	792.41,433 ms TTFT	964.32,869 ms TTFT	Over capacity
16K	61.1650 ms TTFT	113.91,104 ms TTFT	161.02,130 ms TTFT	238.43,568 ms TTFT	285.55,873 ms TTFT	Over capacity	Over capacity	Over capacity
32K	44.5785 ms TTFT	87.81,285 ms TTFT	111.82,407 ms TTFT	158.74,032 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
64K	31.41,116 ms TTFT	59.11,647 ms TTFT	70.52,981 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
128K	18.32,375 ms TTFT	36.13,809 ms TTFT	40.56,305 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
248K	11.2Requests queued3,525 ms TTFT	18.3Requests queued5,405 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	1,921.0	—	4,239 ms	—	—	1	—	—
Prefill	16K	1	Valid	1,821.0	—	8,848 ms	—	—	1	—	—
Prefill	32K	1	Valid	1,713.0	—	18,730 ms	—	—	1	—	—
Prefill	64K	1	Valid	1,509.0	—	42,419 ms	—	—	1	—	—
Prefill	128K	1	Valid	1,219.0	—	104,896 ms	—	—	1	—	—
Sustained decode	0	1	Valid	85.1	86.3	167 ms	11.6 ms	1.0	0	2,554	30.0 s
Sustained decode	0	2	Valid	162.5	82.3	226 ms	12.2 ms	2.0	0	4,874	30.0 s
Sustained decode	0	4	Valid	274.6	69.5	379 ms	14.4 ms	4.0	0	8,233	30.0 s
Sustained decode	0	8	Valid	451.4	57.4	602 ms	17.4 ms	8.0	0	13,524	30.0 s
Sustained decode	0	16	Valid	689.8	44.3	941 ms	22.6 ms	16.0	0	20,693	30.0 s
Sustained decode	0	32	Valid	792.4	25.4	1,433 ms	39.4 ms	32.0	0	23,714	29.9 s
Sustained decode	0	64	Valid	964.3	15.3	2,869 ms	65.4 ms	64.0	0	28,846	29.9 s
Sustained decode	0	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	61.1	62.1	650 ms	16.1 ms	1.0	0	1,832	30.0 s
Sustained decode	16K	2	Valid	113.9	57.5	1,104 ms	17.4 ms	2.0	0	3,414	30.0 s
Sustained decode	16K	4	Valid	161.0	40.7	2,130 ms	24.6 ms	4.0	0	4,828	30.0 s
Sustained decode	16K	8	Valid	238.4	29.9	3,568 ms	33.4 ms	8.0	0	7,139	29.9 s
Sustained decode	16K	16	Valid	285.5	16.7	5,873 ms	59.9 ms	16.0	0	8,540	29.9 s
Sustained decode	16K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	32K	1	Valid	44.5	45.6	785 ms	21.9 ms	1.0	0	1,333	30.0 s
Sustained decode	32K	2	Valid	87.8	44.3	1,285 ms	22.6 ms	2.0	0	2,631	30.0 s
Sustained decode	32K	4	Valid	111.8	28.2	2,407 ms	35.4 ms	4.0	0	3,349	29.9 s
Sustained decode	32K	8	Valid	158.7	19.8	4,032 ms	50.5 ms	8.0	0	4,747	29.9 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	31.4	31.7	1,116 ms	31.5 ms	1.0	0	941	30.0 s
Sustained decode	64K	2	Valid	59.1	29.8	1,647 ms	33.6 ms	2.0	0	1,771	30.0 s
Sustained decode	64K	4	Valid	70.5	17.9	2,981 ms	55.9 ms	4.0	0	2,109	29.9 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	18.3	18.6	2,375 ms	53.7 ms	1.0	0	546	29.9 s
Sustained decode	128K	2	Valid	36.1	18.4	3,809 ms	54.4 ms	2.0	0	1,081	29.9 s
Sustained decode	128K	4	Valid	40.5	9.8	6,305 ms	101.6 ms	4.0	0	1,212	29.9 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	11.2	12.4	3,525 ms	80.4 ms	0.9	0	336	29.9 s
Sustained decode	248K	2	Requests queued	18.3	10.9	5,405 ms	91.7 ms	1.7	0	549	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.