Speed test matrix

Nanbeige4.3-33B-FP8 on CMP 170HX 64GB

vllm · 0.32.0rc1.dev249+g18305948a · FP87 contexts (0–416K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K

by @Sandbergllm-inference-bench v0.8.0Measured Oct 31, 2026, 17:19 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. KV cache: int4_per_token_head
# 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 primitive-ai/Nanbeige4.3-33B-FP8 --max-model-len 434176 --gpu-memory-utilization 0.94 --max-num-seqs 128 --enable-prefix-caching --trust-remote-code  --kv-cache-dtype int4_per_token_head --attention-backend TRITON_ATTN

Prefill

prompt tok/s by context length

8K context1,0157,966 ms TTFT

16K context671.023,786 ms TTFT

32K context414.076,765 ms TTFT

64K context234.0270,261 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	32.99,533 ms TTFT	64.29,701 ms TTFT	123.4266 ms TTFT	233.2317 ms TTFT	421.4394 ms TTFT	612.9613 ms TTFT	Over capacity	Over capacity
16K	25.6282 ms TTFT	40.9402 ms TTFT	66.9514 ms TTFT	98.5695 ms TTFT	123.61,074 ms TTFT	Over capacity	Over capacity	Over capacity
32K	20.7Requests queued467 ms TTFT	29.9Requests queued601 ms TTFT	45.9912 ms TTFT	62.11,275 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
64K	14.8Requests queued911 ms TTFT	19.2Requests queued1,092 ms TTFT	28.01,612 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
128K	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
256K	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
416K	Over capacity	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	1,015.0	—	7,966 ms	—	—	1	—	—
Prefill	16K	1	Valid	671.0	—	23,786 ms	—	—	1	—	—
Prefill	32K	1	Valid	414.0	—	76,765 ms	—	—	1	—	—
Prefill	64K	1	Valid	234.0	—	270,261 ms	—	—	1	—	—
Sustained decode	0	1	Valid	32.9	32.9	9,533 ms	30.4 ms	1.0	0	985	30.0 s
Sustained decode	0	2	Valid	64.2	32.1	9,701 ms	31.1 ms	2.0	0	1,924	30.0 s
Sustained decode	0	4	Valid	123.4	31.0	266 ms	32.3 ms	4.0	0	3,700	30.0 s
Sustained decode	0	8	Valid	233.2	29.4	317 ms	34.0 ms	8.0	0	6,992	30.0 s
Sustained decode	0	16	Valid	421.4	26.7	394 ms	37.4 ms	16.0	0	12,640	30.0 s
Sustained decode	0	32	Valid	612.9	19.4	613 ms	51.5 ms	32.0	0	18,368	30.0 s
Sustained decode	0	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	25.6	25.6	282 ms	39.0 ms	1.0	0	766	30.0 s
Sustained decode	16K	2	Valid	40.9	20.5	402 ms	48.8 ms	2.0	0	1,228	30.0 s
Sustained decode	16K	4	Valid	66.9	16.8	514 ms	59.6 ms	4.0	0	2,008	30.0 s
Sustained decode	16K	8	Valid	98.5	12.3	695 ms	81.1 ms	8.0	0	2,952	30.0 s
Sustained decode	16K	16	Valid	123.6	7.7	1,074 ms	129.0 ms	16.0	0	3,696	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	Requests queued	20.7	20.7	467 ms	48.2 ms	1.0	0	620	30.0 s
Sustained decode	32K	2	Requests queued	29.9	15.0	601 ms	66.6 ms	2.0	0	896	30.0 s
Sustained decode	32K	4	Valid	45.9	11.5	912 ms	87.2 ms	4.0	0	1,376	30.0 s
Sustained decode	32K	8	Valid	62.1	7.8	1,275 ms	128.7 ms	8.0	0	1,856	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	Requests queued	14.8	15.0	911 ms	66.6 ms	1.0	0	445	30.0 s
Sustained decode	64K	2	Requests queued	19.2	9.8	1,092 ms	102.5 ms	1.9	0	574	29.9 s
Sustained decode	64K	4	Valid	28.0	7.0	1,612 ms	143.1 ms	4.0	0	836	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	Over capacity	—	—	—	—	0.4	0	0	30.0 s
Sustained decode	128K	2	Over capacity	—	—	—	—	0.4	0	0	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	256K	1	Over capacity	—	—	—	—	0.6	0	0	30.0 s
Sustained decode	256K	2	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	256K	4	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	256K	8	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	256K	16	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	256K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	256K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	256K	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	416K	1	Over capacity	—	—	—	—	0.6	0	0	30.0 s
Sustained decode	416K	2	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	416K	4	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	416K	8	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	416K	16	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	416K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	416K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	416K	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.