Speed test matrix
Qwen3-8B-AWQ on RTX 3090 24GB
vllm · 0.32.0 · AWQ3 contexts (0–32K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K
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
- AWQ
- 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 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 Qwen/Qwen3-8B-AWQ --max-model-len 40960 --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
Prefill
prompt tok/s by context length8K context3,8242,158 ms TTFT
16K context2,6856,089 ms TTFT
32K context1,67719,404 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 | 107.033 ms TTFT | 208.261 ms TTFT | 379.566 ms TTFT | 670.073 ms TTFT | 1,094.1104 ms TTFT | 1,749.9122 ms TTFT | Over capacity | Over capacity |
| 16K | 70.294 ms TTFT | 126.0139 ms TTFT | 177.1189 ms TTFT | 239.6284 ms TTFT | 299.8418 ms TTFT | Over capacity | Over capacity | Over capacity |
| 32K | 51.9173 ms TTFT | 89.3247 ms TTFT | 114.1319 ms TTFT | 143.4445 ms TTFT | 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 | 3,824.0 | — | 2,158 ms | — | — | 1 | — | — |
| Prefill | 16K | 1 | Valid | 2,685.0 | — | 6,089 ms | — | — | 1 | — | — |
| Prefill | 32K | 1 | Valid | 1,677.0 | — | 19,404 ms | — | — | 1 | — | — |
| Sustained decode | 0 | 1 | Valid | 107.0 | 108.2 | 33 ms | 9.2 ms | 1.0 | 0 | 3,211 | 30.0 s |
| Sustained decode | 0 | 2 | Valid | 208.2 | 105.5 | 61 ms | 9.5 ms | 2.0 | 0 | 6,245 | 30.0 s |
| Sustained decode | 0 | 4 | Valid | 379.5 | 97.0 | 66 ms | 10.3 ms | 4.0 | 0 | 11,391 | 30.0 s |
| Sustained decode | 0 | 8 | Valid | 670.0 | 86.7 | 73 ms | 11.5 ms | 8.0 | 0 | 20,096 | 30.0 s |
| Sustained decode | 0 | 16 | Valid | 1,094.1 | 72.0 | 104 ms | 13.9 ms | 16.0 | 0 | 32,816 | 30.0 s |
| Sustained decode | 0 | 32 | Valid | 1,749.9 | 59.1 | 122 ms | 16.9 ms | 32.0 | 0 | 52,480 | 30.0 s |
| Sustained decode | 0 | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 0 | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 1 | Valid | 70.2 | 70.4 | 94 ms | 14.2 ms | 1.0 | 0 | 2,105 | 30.0 s |
| Sustained decode | 16K | 2 | Valid | 126.0 | 63.2 | 139 ms | 15.8 ms | 2.0 | 0 | 3,780 | 30.0 s |
| Sustained decode | 16K | 4 | Valid | 177.1 | 44.4 | 189 ms | 22.5 ms | 4.0 | 0 | 5,312 | 30.0 s |
| Sustained decode | 16K | 8 | Valid | 239.6 | 30.1 | 284 ms | 33.2 ms | 8.0 | 0 | 7,184 | 30.0 s |
| Sustained decode | 16K | 16 | Valid | 299.8 | 18.8 | 418 ms | 53.3 ms | 16.0 | 0 | 8,992 | 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 | 51.9 | 52.0 | 173 ms | 19.2 ms | 1.0 | 0 | 1,557 | 30.0 s |
| Sustained decode | 32K | 2 | Valid | 89.3 | 44.8 | 247 ms | 22.3 ms | 2.0 | 0 | 2,680 | 30.0 s |
| Sustained decode | 32K | 4 | Valid | 114.1 | 28.6 | 319 ms | 35.0 ms | 4.0 | 0 | 3,420 | 30.0 s |
| Sustained decode | 32K | 8 | Valid | 143.4 | 17.9 | 445 ms | 55.8 ms | 8.0 | 0 | 4,296 | 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 | — | — | — | — | — | — | — | — |
Over capacity: not run because this many requests at this context would exceed the server's KV cache.