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 31, 2026, 12:41 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. KV cache: int4_per_token_head
# 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 --kv-cache-dtype int4_per_token_head --attention-backend TRITON_ATTN --max-cudagraph-capture-size 16

Prefill

prompt tok/s by context length

8K context1,2406,729 ms TTFT

16K context1,12214,716 ms TTFT

32K context964.034,092 ms TTFT

64K context756.086,760 ms TTFT

120K context547.0224,623 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	55.1144 ms TTFT	104.0181 ms TTFT	187.9795 ms TTFT	315.3604 ms TTFT	470.71,129 ms TTFT	418.3Requests queued1,473 ms TTFT	Over capacity	Over capacity
16K	51.11,952 ms TTFT	91.62,959 ms TTFT	154.44,832 ms TTFT	235.28,628 ms TTFT	315.116,383 ms TTFT	Over capacity	Over capacity	Over capacity
32K	47.84,665 ms TTFT	81.66,971 ms TTFT	130.911,512 ms TTFT	187.020,643 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity
64K	42.23,959 ms TTFT	67.55,807 ms TTFT	100.89,539 ms TTFT	Over capacity	Over capacity	Over capacity	Over capacity	Over capacity
120K	33.9Requests queued1,795 ms TTFT	47.6Requests queued2,541 ms TTFT	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	1,240.0	—	6,729 ms	—	—	1	—	—
Prefill	16K	1	Valid	1,122.0	—	14,716 ms	—	—	1	—	—
Prefill	32K	1	Valid	964.0	—	34,092 ms	—	—	1	—	—
Prefill	64K	1	Valid	756.0	—	86,760 ms	—	—	1	—	—
Prefill	120K	1	Valid	547.0	—	224,623 ms	—	—	1	—	—
Sustained decode	0	1	Valid	55.1	55.2	144 ms	18.1 ms	1.0	0	1,653	30.0 s
Sustained decode	0	2	Valid	104.0	52.1	181 ms	19.2 ms	2.0	0	3,119	30.0 s
Sustained decode	0	4	Valid	187.9	47.1	795 ms	21.2 ms	4.0	0	5,636	30.0 s
Sustained decode	0	8	Valid	315.3	39.5	604 ms	25.3 ms	8.0	0	9,459	30.0 s
Sustained decode	0	16	Valid	470.7	29.5	1,129 ms	33.9 ms	16.0	0	14,112	30.0 s
Sustained decode	0	32	Requests queued	418.3	20.2	1,473 ms	49.4 ms	21.0	0	12,537	30.0 s
Sustained decode	0	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	0	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	1	Valid	51.1	51.3	1,952 ms	19.5 ms	1.0	0	1,534	30.0 s
Sustained decode	16K	2	Valid	91.6	44.7	2,959 ms	22.4 ms	2.0	0	2,746	30.0 s
Sustained decode	16K	4	Valid	154.4	35.7	4,832 ms	28.0 ms	4.0	0	4,628	30.0 s
Sustained decode	16K	8	Valid	235.2	24.7	8,628 ms	40.4 ms	8.0	0	7,056	30.0 s
Sustained decode	16K	16	Valid	315.1	14.1	16,383 ms	71.1 ms	16.0	0	9,440	30.0 s
Sustained decode	16K	32	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	64	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	16K	128	Over capacity	—	—	—	—	—	—	—	—
Sustained decode	32K	1	Valid	47.8	47.9	4,665 ms	20.9 ms	1.0	0	1,432	30.0 s
Sustained decode	32K	2	Valid	81.6	38.5	6,971 ms	26.1 ms	2.0	0	2,448	30.0 s
Sustained decode	32K	4	Valid	130.9	27.4	11,512 ms	36.6 ms	4.0	0	3,924	30.0 s
Sustained decode	32K	8	Valid	187.0	16.1	20,643 ms	62.1 ms	8.0	0	5,608	30.0 s
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	42.2	42.4	3,959 ms	23.6 ms	1.0	0	1,266	30.0 s
Sustained decode	64K	2	Valid	67.5	32.2	5,807 ms	31.1 ms	2.0	0	2,026	30.0 s
Sustained decode	64K	4	Valid	100.8	21.9	9,539 ms	45.8 ms	4.0	0	3,021	30.0 s
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	33.9	35.4	1,795 ms	28.2 ms	0.9	0	1,017	30.0 s
Sustained decode	120K	2	Requests queued	47.6	25.5	2,541 ms	39.2 ms	1.9	0	1,428	30.0 s
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.