Speed test matrix

granite-4.2-8b-fp8 on RTX 3090 24GB

vllm · 0.32.0 · FP85 contexts (0–120K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 120K

by @Sandbergllm-inference-bench v0.8.0Measured Oct 31, 2026, 09:06 UTC

Hardware

Type: Discrete GPU
GPU: RTX 3090
VRAM: 24 GB
CPU: Intel Core i7-9700K (8C)
RAM: 32 GB
OS: Ubuntu 24.04.5

Engine

Engine: vllm
Version: 0.32.0
Quantization: FP8
Backend: cuda
Repository: github.com/vllm-project/vllm
Build: vllm/vllm-openai:v0.32.0-x86_64

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: fp8
# 1x RTX 3090 24GB (Zagitta/gpu-unlock), 420 W limit, PCIe Gen3 x16 | Intel Core i7-9700K 8C | 32 GB DDR4 | Ubuntu 24.04.5, NVIDIA driver 580.178.04
vllm serve ibm-granite/granite-4.2-8b-fp8 --max-model-len 131072 --gpu-memory-utilization 0.94 --max-num-seqs 128 --enable-prefix-caching --trust-remote-code  --kv-cache-dtype fp8

Prefill

prompt tok/s by context length

8K context6,0991,351 ms TTFT

16K context5,2573,103 ms TTFT

32K context4,1217,879 ms TTFT

64K context2,88622,443 ms TTFT

120K context1,87364,767 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	83.926 ms TTFT	164.630 ms TTFT	318.835 ms TTFT	603.040 ms TTFT	1,105.256 ms TTFT	Over capacity	Over capacity	Over capacity
16K	74.456 ms TTFT	145.480 ms TTFT	257.1113 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
32K	66.990 ms TTFT	129.6130 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
64K	55.8182 ms TTFT	105.2267 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
120K	43.3368 ms TTFT	Over capacity	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	6,099.0	—	1,351 ms	—	—	1	—	—
Prefill	16K	1	Valid	5,257.0	—	3,103 ms	—	—	1	—	—
Prefill	32K	1	Valid	4,121.0	—	7,879 ms	—	—	1	—	—
Prefill	64K	1	Valid	2,886.0	—	22,443 ms	—	—	1	—	—
Prefill	120K	1	Valid	1,873.0	—	64,767 ms	—	—	1	—	—
Sustained decode	0	1	Valid	83.9	84.1	26 ms	11.9 ms	1.0	0	2,516	30.0 s
Sustained decode	0	2	Valid	164.6	82.6	30 ms	12.1 ms	2.0	0	4,936	30.0 s
Sustained decode	0	4	Valid	318.8	80.3	35 ms	12.5 ms	4.0	0	9,560	30.0 s
Sustained decode	0	8	Valid	603.0	76.4	40 ms	13.1 ms	8.0	0	18,091	30.0 s
Sustained decode	0	16	Valid	1,105.2	70.7	56 ms	14.1 ms	16.0	0	33,152	30.0 s
Sustained decode	0	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	74.4	74.5	56 ms	13.4 ms	1.0	0	2,231	30.0 s
Sustained decode	16K	2	Valid	145.4	72.8	80 ms	13.7 ms	2.0	0	4,361	30.0 s
Sustained decode	16K	4	Valid	257.1	64.5	113 ms	15.5 ms	4.0	0	7,711	30.0 s
Sustained decode	16K	8	Over capacity	—	—	—	—	—	—	—	—
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	66.9	67.0	90 ms	14.9 ms	1.0	0	2,006	30.0 s
Sustained decode	32K	2	Valid	129.6	64.9	130 ms	15.4 ms	2.0	0	3,886	30.0 s
Sustained decode	32K	4	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	32K	8	Over capacity	—	—	—	—	—	—	—	—
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	55.8	55.9	182 ms	17.9 ms	1.0	0	1,674	30.0 s
Sustained decode	64K	2	Valid	105.2	52.7	267 ms	19.0 ms	2.0	0	3,156	30.0 s
Sustained decode	64K	4	Over capacity	—	—	—	—	—	—	—	—
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	43.3	43.3	368 ms	23.1 ms	1.0	0	1,298	30.0 s
Sustained decode	120K	2	Over capacity	—	—	—	—	—	—	—	—
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.