Speed test matrix

granite-4.2-3b-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 30, 2026, 09:52 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
# 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-4.2-3b-fp8 --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 context12.3K668 ms TTFT

16K context9,9951,633 ms TTFT

32K context7,4464,361 ms TTFT

64K context4,89913,222 ms TTFT

120K context3,04139,885 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	159.916 ms TTFT	298.818 ms TTFT	554.122 ms TTFT	960.229 ms TTFT	1,601.645 ms TTFT	Over capacity	Over capacity	Over capacity
16K	127.955 ms TTFT	214.576 ms TTFT	370.1116 ms TTFT	489.4200 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
32K	107.9103 ms TTFT	166.0135 ms TTFT	272.1203 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
64K	81.8193 ms TTFT	114.0253 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
120K	57.2386 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	12.3K	—	668 ms	—	—	1	—	—
Prefill	16K	1	Valid	9,995.0	—	1,633 ms	—	—	1	—	—
Prefill	32K	1	Valid	7,446.0	—	4,361 ms	—	—	1	—	—
Prefill	64K	1	Valid	4,899.0	—	13,222 ms	—	—	1	—	—
Prefill	120K	1	Valid	3,041.0	—	39,885 ms	—	—	1	—	—
Sustained decode	0	1	Valid	159.9	161.3	16 ms	6.2 ms	1.0	0	4,798	30.0 s
Sustained decode	0	2	Valid	298.8	151.9	18 ms	6.6 ms	2.0	0	8,962	30.0 s
Sustained decode	0	4	Valid	554.1	141.9	22 ms	7.0 ms	4.0	0	16,623	30.0 s
Sustained decode	0	8	Valid	960.2	124.8	29 ms	8.0 ms	8.0	0	28,800	30.0 s
Sustained decode	0	16	Valid	1,601.6	106.2	45 ms	9.4 ms	16.0	0	48,048	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	127.9	128.3	55 ms	7.8 ms	1.0	0	3,837	30.0 s
Sustained decode	16K	2	Valid	214.5	107.8	76 ms	9.3 ms	2.0	0	6,434	30.0 s
Sustained decode	16K	4	Valid	370.1	93.2	116 ms	10.7 ms	4.0	0	11,100	30.0 s
Sustained decode	16K	8	Valid	489.4	62.2	200 ms	16.1 ms	8.0	0	14,680	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	107.9	108.1	103 ms	9.3 ms	1.0	0	3,236	30.0 s
Sustained decode	32K	2	Valid	166.0	83.2	135 ms	12.0 ms	2.0	0	4,978	30.0 s
Sustained decode	32K	4	Valid	272.1	68.0	203 ms	14.7 ms	4.0	0	8,160	30.0 s
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	81.8	82.0	193 ms	12.2 ms	1.0	0	2,453	30.0 s
Sustained decode	64K	2	Valid	114.0	57.1	253 ms	17.5 ms	2.0	0	3,418	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	57.2	57.4	386 ms	17.4 ms	1.0	0	1,717	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.