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
- AMD Ryzen 9 7900X (12C)
- 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), 350 W limit (VBIOS max), PCIe Gen4 x16 | AMD Ryzen 9 7900X 12C | 32 GB DDR5 | Ubuntu 24.04.5, NVIDIA driver 595.99.02 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 context3,3842,438 ms TTFT
16K context2,3536,947 ms TTFT
32K context1,47322,089 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 | 104.129 ms TTFT | 201.934 ms TTFT | 366.836 ms TTFT | 643.448 ms TTFT | 1,041.177 ms TTFT | 1,663.1108 ms TTFT | Over capacity | Over capacity |
| 16K | 67.782 ms TTFT | 119.3123 ms TTFT | 166.8184 ms TTFT | 222.6227 ms TTFT | 271.7340 ms TTFT | Over capacity | Over capacity | Over capacity |
| 32K | 49.9149 ms TTFT | 84.0207 ms TTFT | 106.1285 ms TTFT | 131.5356 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 | 3,384.0 | — | 2,438 ms | — | — | 1 | — | — |
| Prefill | 16K | 1 | Valid | 2,353.0 | — | 6,947 ms | — | — | 1 | — | — |
| Prefill | 32K | 1 | Valid | 1,473.0 | — | 22,089 ms | — | — | 1 | — | — |
| Sustained decode | 0 | 1 | Valid | 104.1 | 105.3 | 29 ms | 9.5 ms | 1.0 | 0 | 3,123 | 30.0 s |
| Sustained decode | 0 | 2 | Valid | 201.9 | 102.4 | 34 ms | 9.8 ms | 2.0 | 0 | 6,056 | 30.0 s |
| Sustained decode | 0 | 4 | Valid | 366.8 | 93.9 | 36 ms | 10.7 ms | 4.0 | 0 | 11,000 | 30.0 s |
| Sustained decode | 0 | 8 | Valid | 643.4 | 83.4 | 48 ms | 12.0 ms | 8.0 | 0 | 19,296 | 30.0 s |
| Sustained decode | 0 | 16 | Valid | 1,041.1 | 68.6 | 77 ms | 14.6 ms | 16.0 | 0 | 31,232 | 30.0 s |
| Sustained decode | 0 | 32 | Valid | 1,663.1 | 56.1 | 108 ms | 17.8 ms | 32.0 | 0 | 49,888 | 30.0 s |
| Sustained decode | 0 | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 0 | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 1 | Valid | 67.7 | 67.9 | 82 ms | 14.7 ms | 1.0 | 0 | 2,030 | 30.0 s |
| Sustained decode | 16K | 2 | Valid | 119.3 | 59.9 | 123 ms | 16.7 ms | 2.0 | 0 | 3,578 | 30.0 s |
| Sustained decode | 16K | 4 | Valid | 166.8 | 41.9 | 184 ms | 23.9 ms | 4.0 | 0 | 5,000 | 30.0 s |
| Sustained decode | 16K | 8 | Valid | 222.6 | 27.9 | 227 ms | 35.8 ms | 8.0 | 0 | 6,672 | 30.0 s |
| Sustained decode | 16K | 16 | Valid | 271.7 | 17.0 | 340 ms | 58.8 ms | 16.0 | 0 | 8,144 | 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 | 49.9 | 50.0 | 149 ms | 20.0 ms | 1.0 | 0 | 1,497 | 30.0 s |
| Sustained decode | 32K | 2 | Valid | 84.0 | 42.1 | 207 ms | 23.7 ms | 2.0 | 0 | 2,520 | 30.0 s |
| Sustained decode | 32K | 4 | Valid | 106.1 | 26.6 | 285 ms | 37.6 ms | 4.0 | 0 | 3,180 | 30.0 s |
| Sustained decode | 32K | 8 | Valid | 131.5 | 16.5 | 356 ms | 60.7 ms | 8.0 | 0 | 3,944 | 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.