Speed test matrix

Tess-4-27B-NVFP4 on CMP 170HX 64GB

vllm · 0.32.0rc1.dev249+g18305948a · NVFP46 contexts (0–248K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 128K

by @Sandbergllm-inference-bench v0.8.0Measured Oct 28, 2026, 21:49 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: NVFP4
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 migtissera/Tess-4-27B-NVFP4 --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 context1,8374,434 ms TTFT

16K context1,7299,324 ms TTFT

32K context1,63019,678 ms TTFT

64K context1,45044,163 ms TTFT

128K context1,189107,526 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	48.6178 ms TTFT	94.6304 ms TTFT	181.4328 ms TTFT	341.1580 ms TTFT	554.1905 ms TTFT	860.71,595 ms TTFT	1,101.02,904 ms TTFT	Over capacity
16K	46.1410 ms TTFT	86.5582 ms TTFT	156.6879 ms TTFT	266.91,582 ms TTFT	386.42,987 ms TTFT	Over capacity	Over capacity	Over capacity
32K	44.1694 ms TTFT	79.91,016 ms TTFT	137.81,595 ms TTFT	219.12,747 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
64K	40.6861 ms TTFT	69.71,132 ms TTFT	111.21,713 ms TTFT	161.42,850 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
128K	35.0968 ms TTFT	54.71,228 ms TTFT	80.41,786 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
248K	26.0Requests queued2,642 ms TTFT	35.1Requests queued3,374 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	1,837.0	—	4,434 ms	—	—	1	—	—
Prefill	16K	1	Valid	1,729.0	—	9,324 ms	—	—	1	—	—
Prefill	32K	1	Valid	1,630.0	—	19,678 ms	—	—	1	—	—
Prefill	64K	1	Valid	1,450.0	—	44,163 ms	—	—	1	—	—
Prefill	128K	1	Valid	1,189.0	—	107,526 ms	—	—	1	—	—
Sustained decode	0	1	Valid	48.6	48.5	178 ms	20.6 ms	1.0	0	1,456	30.0 s
Sustained decode	0	2	Valid	94.6	47.4	304 ms	21.1 ms	2.0	0	2,838	30.0 s
Sustained decode	0	4	Valid	181.4	45.4	328 ms	22.0 ms	4.0	0	5,440	30.0 s
Sustained decode	0	8	Valid	341.1	42.8	580 ms	23.4 ms	8.0	0	10,232	30.0 s
Sustained decode	0	16	Valid	554.1	34.7	905 ms	28.8 ms	16.0	0	16,624	30.0 s
Sustained decode	0	32	Valid	860.7	26.9	1,595 ms	37.1 ms	32.0	0	25,824	30.0 s
Sustained decode	0	64	Valid	1,101.0	17.2	2,904 ms	58.3 ms	64.0	0	33,024	30.0 s
Sustained decode	0	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	46.1	46.2	410 ms	21.7 ms	1.0	0	1,382	30.0 s
Sustained decode	16K	2	Valid	86.5	43.1	582 ms	23.2 ms	2.0	0	2,594	30.0 s
Sustained decode	16K	4	Valid	156.6	38.7	879 ms	25.8 ms	4.0	0	4,696	30.0 s
Sustained decode	16K	8	Valid	266.9	32.5	1,582 ms	30.8 ms	8.0	0	8,008	30.0 s
Sustained decode	16K	16	Valid	386.4	22.7	2,987 ms	44.1 ms	16.0	0	11,584	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	44.1	44.2	694 ms	22.6 ms	1.0	0	1,322	30.0 s
Sustained decode	32K	2	Valid	79.9	39.7	1,016 ms	25.2 ms	2.0	0	2,396	30.0 s
Sustained decode	32K	4	Valid	137.8	33.8	1,595 ms	29.6 ms	4.0	0	4,132	30.0 s
Sustained decode	32K	8	Valid	219.1	26.0	2,747 ms	38.4 ms	8.0	0	6,568	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	40.6	40.7	861 ms	24.6 ms	1.0	0	1,217	30.0 s
Sustained decode	64K	2	Valid	69.7	34.7	1,132 ms	28.8 ms	2.0	0	2,092	30.0 s
Sustained decode	64K	4	Valid	111.2	27.3	1,713 ms	36.6 ms	4.0	0	3,336	30.0 s
Sustained decode	64K	8	Valid	161.4	19.2	2,850 ms	52.1 ms	8.0	0	4,840	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	35.0	35.2	968 ms	28.4 ms	1.0	0	1,050	30.0 s
Sustained decode	128K	2	Valid	54.7	27.4	1,228 ms	36.5 ms	2.0	0	1,642	30.0 s
Sustained decode	128K	4	Valid	80.4	19.9	1,786 ms	50.1 ms	4.0	0	2,408	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	26.0	28.0	2,642 ms	35.7 ms	0.9	0	781	30.0 s
Sustained decode	248K	2	Requests queued	35.1	19.4	3,374 ms	51.6 ms	1.8	0	1,052	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.