Speed test matrix

granite-4.2-8b-nvfp4 on CMP 170HX 64GB

vllm · 0.32.0rc1.dev249+g18305948a · NVFP45 contexts (0–120K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 120K

by @Sandbergllm-inference-bench v0.8.0Measured Oct 29, 2026, 02:03 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 ibm-granite/granite-4.2-8b-nvfp4 --max-model-len 131072 --gpu-memory-utilization 0.94 --max-num-seqs 128 --enable-prefix-caching --trust-remote-code

Prefill

prompt tok/s by context length

8K context5,1231,600 ms TTFT

16K context4,4953,607 ms TTFT

32K context3,7098,699 ms TTFT

64K context2,71623,701 ms TTFT

120K context1,84865,222 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	127.159 ms TTFT	237.649 ms TTFT	440.064 ms TTFT	775.883 ms TTFT	1,221.8141 ms TTFT	1,825.1219 ms TTFT	Over capacity	Over capacity
16K	98.8195 ms TTFT	162.4242 ms TTFT	248.1300 ms TTFT	338.4383 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
32K	80.4241 ms TTFT	119.9364 ms TTFT	164.2538 ms TTFT	204.5841 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
64K	58.9479 ms TTFT	77.3634 ms TTFT	97.5920 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
120K	39.8862 ms TTFT	46.91,041 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity

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	5,123.0	—	1,600 ms	—	—	1	—	—
Prefill	16K	1	Valid	4,495.0	—	3,607 ms	—	—	1	—	—
Prefill	32K	1	Valid	3,709.0	—	8,699 ms	—	—	1	—	—
Prefill	64K	1	Valid	2,716.0	—	23,701 ms	—	—	1	—	—
Prefill	120K	1	Valid	1,848.0	—	65,222 ms	—	—	1	—	—
Sustained decode	0	1	Valid	127.1	128.3	59 ms	7.8 ms	1.0	0	3,811	30.0 s
Sustained decode	0	2	Valid	237.6	120.6	49 ms	8.3 ms	2.0	0	7,126	30.0 s
Sustained decode	0	4	Valid	440.0	112.8	64 ms	8.9 ms	4.0	0	13,197	30.0 s
Sustained decode	0	8	Valid	775.8	100.9	83 ms	9.9 ms	8.0	0	23,270	30.0 s
Sustained decode	0	16	Valid	1,221.8	80.8	141 ms	12.4 ms	16.0	0	36,656	30.0 s
Sustained decode	0	32	Valid	1,825.1	60.9	219 ms	16.4 ms	32.0	0	54,756	30.0 s
Sustained decode	0	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	98.8	99.2	195 ms	10.1 ms	1.0	0	2,964	30.0 s
Sustained decode	16K	2	Valid	162.4	81.7	242 ms	12.2 ms	2.0	0	4,870	30.0 s
Sustained decode	16K	4	Valid	248.1	62.4	300 ms	16.0 ms	4.0	0	7,444	30.0 s
Sustained decode	16K	8	Valid	338.4	42.4	383 ms	23.6 ms	8.0	0	10,152	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	80.4	80.8	241 ms	12.4 ms	1.0	0	2,413	30.0 s
Sustained decode	32K	2	Valid	119.9	60.2	364 ms	16.6 ms	2.0	0	3,596	30.0 s
Sustained decode	32K	4	Valid	164.2	41.2	538 ms	24.3 ms	4.0	0	4,924	30.0 s
Sustained decode	32K	8	Valid	204.5	25.5	841 ms	39.2 ms	8.0	0	6,128	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	58.9	59.1	479 ms	16.9 ms	1.0	0	1,766	30.0 s
Sustained decode	64K	2	Valid	77.3	38.8	634 ms	25.8 ms	2.0	0	2,319	30.0 s
Sustained decode	64K	4	Valid	97.5	24.4	920 ms	40.9 ms	4.0	0	2,924	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	120K	1	Valid	39.8	40.0	862 ms	25.0 ms	1.0	0	1,193	30.0 s
Sustained decode	120K	2	Valid	46.9	23.6	1,041 ms	42.4 ms	2.0	0	1,406	30.0 s
Sustained decode	120K	4	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	120K	8	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	120K	16	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	120K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	120K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	120K	128	Over capacity	—	—	—	—	—	—	—	—

Over capacity: not run because this many requests at this context would exceed the server's KV cache.