Speed test matrix

NVIDIA-Nemotron-Nano-9B-v2-quantized.w4a16 on RTX 3060 12GB

vllm · 0.32.0 · W4A165 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, 22:15 UTC

Hardware

Type: Discrete GPU
GPU: RTX 3060
VRAM: 12 GB
CPU: Intel Xeon E5-2680 v4 (14C)
RAM: 32 GB
OS: Ubuntu 24.04.5

Engine

Engine: vllm
Version: 0.32.0
Quantization: W4A16
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 3060 12GB, stock 170 W limit, PCIe Gen3 x16 | Intel Xeon E5-2680 v4 14C | 32 GB DDR4-2400 | Ubuntu 24.04.5, NVIDIA driver 580.178.04
vllm serve RedHatAI/NVIDIA-Nemotron-Nano-9B-v2-quantized.w4a16 --max-model-len 131072 --gpu-memory-utilization 0.94 --max-num-seqs 128 --enable-prefix-caching --trust-remote-code  --max-cudagraph-capture-size 16

Prefill

prompt tok/s by context length

8K context1,3376,240 ms TTFT

16K context1,28612,838 ms TTFT

32K context1,23726,578 ms TTFT

64K context1,14157,463 ms TTFT

120K context1,005122,223 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	55.7130 ms TTFT	105.1658 ms TTFT	190.3305 ms TTFT	320.2597 ms TTFT	480.51,105 ms TTFT	Over capacity	Over capacity	Over capacity
16K	53.0717 ms TTFT	95.81,078 ms TTFT	161.81,726 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
32K	50.5256 ms TTFT	88.0357 ms TTFT	140.9496 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
64K	46.01,615 ms TTFT	75.62,309 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
120K	38.8Requests queued1,628 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity

Requests queued: the server could not run all requests at once, so some waited in a queue. 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	1,337.0	—	6,240 ms	—	—	1	—	—
Prefill	16K	1	Valid	1,286.0	—	12,838 ms	—	—	1	—	—
Prefill	32K	1	Valid	1,237.0	—	26,578 ms	—	—	1	—	—
Prefill	64K	1	Valid	1,141.0	—	57,463 ms	—	—	1	—	—
Prefill	120K	1	Valid	1,005.0	—	122,223 ms	—	—	1	—	—
Sustained decode	0	1	Valid	55.7	55.8	130 ms	17.9 ms	1.0	0	1,671	30.0 s
Sustained decode	0	2	Valid	105.1	52.7	658 ms	19.0 ms	2.0	0	3,154	30.0 s
Sustained decode	0	4	Valid	190.3	47.6	305 ms	21.0 ms	4.0	0	5,708	30.0 s
Sustained decode	0	8	Valid	320.2	40.1	597 ms	24.9 ms	8.0	0	9,600	30.0 s
Sustained decode	0	16	Valid	480.5	30.1	1,105 ms	33.2 ms	16.0	0	14,417	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	53.0	53.1	717 ms	18.8 ms	1.0	0	1,590	30.0 s
Sustained decode	16K	2	Valid	95.8	47.5	1,078 ms	21.0 ms	2.0	0	2,874	30.0 s
Sustained decode	16K	4	Valid	161.8	39.5	1,726 ms	25.3 ms	4.0	0	4,851	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	50.5	50.5	256 ms	19.8 ms	1.0	0	1,513	30.0 s
Sustained decode	32K	2	Valid	88.0	44.0	357 ms	22.7 ms	2.0	0	2,638	30.0 s
Sustained decode	32K	4	Valid	140.9	35.2	496 ms	28.4 ms	4.0	0	4,223	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	46.0	46.2	1,615 ms	21.7 ms	1.0	0	1,379	30.0 s
Sustained decode	64K	2	Valid	75.6	37.2	2,309 ms	26.9 ms	2.0	0	2,266	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	Requests queued	38.8	40.2	1,628 ms	24.9 ms	1.0	0	1,162	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	—	—	—	—	—	—	—	—

Requests queued: the server could not run all requests at once, so some waited in a queue. Over capacity: not run because this many requests at this context would exceed the server's KV cache.