Speed test matrix

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

vllm · 0.32.0 · FP8 dynamic5 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, 14:36 UTC

Hardware

Type: Discrete GPU
GPU: RTX 3090
VRAM: 24 GB
CPU: AMD Ryzen 9 7900X (12C)
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. KV cache: fp8
# 1x RTX 3090 24GB (Zagitta/gpu-unlock), 350 W limit (VBIOS max), PCIe Gen4 x16 | AMD Ryzen 9 7900X 12C | 32 GB DDR5 | Ubuntu 24.04.5, NVIDIA driver 595.99.02
vllm serve RedHatAI/gemma-3-12b-it-FP8-dynamic --max-model-len 131072 --gpu-memory-utilization 0.94 --max-num-seqs 128 --enable-prefix-caching --trust-remote-code --language-model-only --kv-cache-dtype fp8

Prefill

prompt tok/s by context length

8K context4,3701,833 ms TTFT

16K context4,1913,787 ms TTFT

32K context3,9288,043 ms TTFT

64K context3,49318,043 ms TTFT

120K context2,93740,193 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	57.635 ms TTFT	112.645 ms TTFT	217.249 ms TTFT	410.154 ms TTFT	745.170 ms TTFT	Over capacity	Over capacity	Over capacity
16K	55.561 ms TTFT	107.182 ms TTFT	199.0117 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
32K	53.897 ms TTFT	102.5139 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
64K	50.4171 ms TTFT	93.8239 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
120K	45.8346 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	4,370.0	—	1,833 ms	—	—	1	—	—
Prefill	16K	1	Valid	4,191.0	—	3,787 ms	—	—	1	—	—
Prefill	32K	1	Valid	3,928.0	—	8,043 ms	—	—	1	—	—
Prefill	64K	1	Valid	3,493.0	—	18,043 ms	—	—	1	—	—
Prefill	120K	1	Valid	2,937.0	—	40,193 ms	—	—	1	—	—
Sustained decode	0	1	Valid	57.6	57.7	35 ms	17.3 ms	1.0	0	1,728	30.0 s
Sustained decode	0	2	Valid	112.6	56.5	45 ms	17.7 ms	2.0	0	3,378	30.0 s
Sustained decode	0	4	Valid	217.2	54.7	49 ms	18.3 ms	4.0	0	6,512	30.0 s
Sustained decode	0	8	Valid	410.1	51.9	54 ms	19.3 ms	8.0	0	12,304	30.0 s
Sustained decode	0	16	Valid	745.1	47.6	70 ms	21.0 ms	16.0	0	22,352	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	55.5	55.5	61 ms	18.0 ms	1.0	0	1,665	30.0 s
Sustained decode	16K	2	Valid	107.1	53.6	82 ms	18.6 ms	2.0	0	3,214	30.0 s
Sustained decode	16K	4	Valid	199.0	49.8	117 ms	20.1 ms	4.0	0	5,968	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	53.8	53.8	97 ms	18.6 ms	1.0	0	1,613	30.0 s
Sustained decode	32K	2	Valid	102.5	51.3	139 ms	19.5 ms	2.0	0	3,074	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	50.4	50.4	171 ms	19.8 ms	1.0	0	1,511	30.0 s
Sustained decode	64K	2	Valid	93.8	46.9	239 ms	21.3 ms	2.0	0	2,812	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	45.8	45.8	346 ms	21.8 ms	1.0	0	1,372	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.