Speed test matrix

gemma-3-12b-it-FP8-dynamic on RTX 3090 24GB

vllm · 0.32.0 · FP8 dynamic5 contexts (0–72K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 72K

by @Sandbergllm-inference-bench v0.8.0Measured Oct 30, 2026, 05:18 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 dynamic
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
# 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 RedHatAI/gemma-3-12b-it-FP8-dynamic --max-model-len 81920 --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 context4,9681,612 ms TTFT

16K context4,6883,385 ms TTFT

32K context4,4557,091 ms TTFT

64K context4,06315,512 ms TTFT

72K context3,98217,805 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	54.432 ms TTFT	107.536 ms TTFT	211.248 ms TTFT	372.456 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
16K	50.4128 ms TTFT	95.1111 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
32K	47.6178 ms TTFT	86.2204 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
64K	42.8306 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
72K	41.7353 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	4,968.0	—	1,612 ms	—	—	1	—	—
Prefill	16K	1	Valid	4,688.0	—	3,385 ms	—	—	1	—	—
Prefill	32K	1	Valid	4,455.0	—	7,091 ms	—	—	1	—	—
Prefill	64K	1	Valid	4,063.0	—	15,512 ms	—	—	1	—	—
Prefill	72K	1	Valid	3,982.0	—	17,805 ms	—	—	1	—	—
Sustained decode	0	1	Valid	54.4	55.1	32 ms	18.2 ms	1.0	0	1,630	30.0 s
Sustained decode	0	2	Valid	107.5	54.5	36 ms	18.3 ms	2.0	0	3,224	30.0 s
Sustained decode	0	4	Valid	211.2	53.6	48 ms	18.7 ms	4.0	0	6,336	30.0 s
Sustained decode	0	8	Valid	372.4	47.9	56 ms	20.9 ms	8.0	0	11,168	30.0 s
Sustained decode	0	16	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	50.4	50.5	128 ms	19.8 ms	1.0	0	1,512	30.0 s
Sustained decode	16K	2	Valid	95.1	47.6	111 ms	21.0 ms	2.0	0	2,852	30.0 s
Sustained decode	16K	4	Over capacity	—	—	—	—	—	—	—	—
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	47.6	47.6	178 ms	21.0 ms	1.0	0	1,427	30.0 s
Sustained decode	32K	2	Valid	86.2	43.2	204 ms	23.2 ms	2.0	0	2,586	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	42.8	42.8	306 ms	23.3 ms	1.0	0	1,282	30.0 s
Sustained decode	64K	2	Over capacity	—	—	—	—	—	—	—	—
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	72K	1	Valid	41.7	41.8	353 ms	23.9 ms	1.0	0	1,251	30.0 s
Sustained decode	72K	2	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	72K	4	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	72K	8	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	72K	16	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	72K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	72K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	72K	128	Over capacity	—	—	—	—	—	—	—	—

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