Speed test matrix

Tess-4-27B-FP8 on CMP 170HX 64GB

vllm · 0.32.0rc1.dev249+g18305948a · FP86 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, 03:12 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: FP8
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 lflare/Tess-4-27B-FP8 --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 context2,2103,685 ms TTFT

16K context2,0737,773 ms TTFT

32K context1,93716,566 ms TTFT

64K context1,68637,976 ms TTFT

128K context1,34195,359 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	33.2197 ms TTFT	64.7208 ms TTFT	125.7362 ms TTFT	240.0537 ms TTFT	455.8819 ms TTFT	672.01,338 ms TTFT	Over capacity	Over capacity
16K	32.0377 ms TTFT	60.7557 ms TTFT	113.1885 ms TTFT	200.21,542 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
32K	31.0640 ms TTFT	57.4909 ms TTFT	102.91,423 ms TTFT	172.12,484 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
64K	29.2855 ms TTFT	51.61,153 ms TTFT	87.41,687 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
128K	26.11,054 ms TTFT	43.01,263 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
248K	20.3Requests queued2,878 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	2,210.0	—	3,685 ms	—	—	1	—	—
Prefill	16K	1	Valid	2,073.0	—	7,773 ms	—	—	1	—	—
Prefill	32K	1	Valid	1,937.0	—	16,566 ms	—	—	1	—	—
Prefill	64K	1	Valid	1,686.0	—	37,976 ms	—	—	1	—	—
Prefill	128K	1	Valid	1,341.0	—	95,359 ms	—	—	1	—	—
Sustained decode	0	1	Valid	33.2	33.2	197 ms	30.1 ms	1.0	0	996	30.0 s
Sustained decode	0	2	Valid	64.7	32.4	208 ms	30.9 ms	2.0	0	1,940	30.0 s
Sustained decode	0	4	Valid	125.7	31.5	362 ms	31.8 ms	4.0	0	3,768	30.0 s
Sustained decode	0	8	Valid	240.0	30.1	537 ms	33.3 ms	8.0	0	7,200	30.0 s
Sustained decode	0	16	Valid	455.8	28.6	819 ms	35.0 ms	16.0	0	13,680	30.0 s
Sustained decode	0	32	Valid	672.0	21.0	1,338 ms	47.6 ms	32.0	0	20,160	30.0 s
Sustained decode	0	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	32.0	32.0	377 ms	31.2 ms	1.0	0	959	30.0 s
Sustained decode	16K	2	Valid	60.7	30.3	557 ms	33.0 ms	2.0	0	1,822	30.0 s
Sustained decode	16K	4	Valid	113.1	28.0	885 ms	35.7 ms	4.0	0	3,388	30.0 s
Sustained decode	16K	8	Valid	200.2	24.4	1,542 ms	40.9 ms	8.0	0	6,000	30.0 s
Sustained decode	16K	16	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	32K	1	Valid	31.0	31.0	640 ms	32.2 ms	1.0	0	929	30.0 s
Sustained decode	32K	2	Valid	57.4	28.6	909 ms	35.0 ms	2.0	0	1,722	30.0 s
Sustained decode	32K	4	Valid	102.9	25.3	1,423 ms	39.5 ms	4.0	0	3,084	30.0 s
Sustained decode	32K	8	Valid	172.1	20.6	2,484 ms	48.6 ms	8.0	0	5,160	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	29.2	29.3	855 ms	34.2 ms	1.0	0	875	30.0 s
Sustained decode	64K	2	Valid	51.6	25.7	1,153 ms	38.9 ms	2.0	0	1,548	30.0 s
Sustained decode	64K	4	Valid	87.4	21.5	1,687 ms	46.6 ms	4.0	0	2,620	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	26.1	26.2	1,054 ms	38.1 ms	1.0	0	784	30.0 s
Sustained decode	128K	2	Valid	43.0	21.6	1,263 ms	46.3 ms	2.0	0	1,290	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	20.3	22.1	2,878 ms	45.3 ms	0.9	0	610	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.