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, 07:34 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. 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 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 context4,8261,660 ms TTFT

16K context4,6503,413 ms TTFT

32K context4,3187,316 ms TTFT

64K context3,88016,248 ms TTFT

120K context3,25336,293 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	58.634 ms TTFT	114.336 ms TTFT	220.351 ms TTFT	416.757 ms TTFT	759.494 ms TTFT	Over capacity	Over capacity	Over capacity
16K	56.578 ms TTFT	109.4169 ms TTFT	203.9167 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
32K	54.7181 ms TTFT	104.9184 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
64K	51.2270 ms TTFT	96.6374 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
120K	46.4523 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	4,826.0	—	1,660 ms	—	—	1	—	—
Prefill	16K	1	Valid	4,650.0	—	3,413 ms	—	—	1	—	—
Prefill	32K	1	Valid	4,318.0	—	7,316 ms	—	—	1	—	—
Prefill	64K	1	Valid	3,880.0	—	16,248 ms	—	—	1	—	—
Prefill	120K	1	Valid	3,253.0	—	36,293 ms	—	—	1	—	—
Sustained decode	0	1	Valid	58.6	58.6	34 ms	17.1 ms	1.0	0	1,757	30.0 s
Sustained decode	0	2	Valid	114.3	57.4	36 ms	17.4 ms	2.0	0	3,428	30.0 s
Sustained decode	0	4	Valid	220.3	55.5	51 ms	18.0 ms	4.0	0	6,608	30.0 s
Sustained decode	0	8	Valid	416.7	52.8	57 ms	19.0 ms	8.0	0	12,496	30.0 s
Sustained decode	0	16	Valid	759.4	48.5	94 ms	20.6 ms	16.0	0	22,768	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	56.5	56.5	78 ms	17.7 ms	1.0	0	1,694	30.0 s
Sustained decode	16K	2	Valid	109.4	54.7	169 ms	18.3 ms	2.0	0	3,280	30.0 s
Sustained decode	16K	4	Valid	203.9	51.1	167 ms	19.6 ms	4.0	0	6,116	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	54.7	54.7	181 ms	18.3 ms	1.0	0	1,640	30.0 s
Sustained decode	32K	2	Valid	104.9	52.5	184 ms	19.0 ms	2.0	0	3,146	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	51.2	51.2	270 ms	19.5 ms	1.0	0	1,534	30.0 s
Sustained decode	64K	2	Valid	96.6	48.4	374 ms	20.7 ms	2.0	0	2,896	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	46.4	46.5	523 ms	21.5 ms	1.0	0	1,391	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.