Speed test matrix
Qwen3-Coder-30B-A3B-Instruct-AWQ-4bit on RTX 3090 24GB
vllm · 0.32.0 · AWQ 4-bit4 contexts (0–64K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K
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 4-bit
- 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: fp8 # 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 cyankiwi/Qwen3-Coder-30B-A3B-Instruct-AWQ-4bit --max-model-len 73728 --gpu-memory-utilization 0.94 --max-num-seqs 128 --enable-prefix-caching --trust-remote-code --kv-cache-dtype fp8
Prefill
prompt tok/s by context length8K context8,583961 ms TTFT
16K context6,7002,440 ms TTFT
32K context4,5237,196 ms TTFT
64K context2,71123,950 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 | 175.027 ms TTFT | 290.531 ms TTFT | 472.731 ms TTFT | 753.138 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity |
| 16K | Looped | Looped | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity |
| 32K | 130.161 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity |
| 64K | 104.5113 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity |
Looped: the model kept repeating the same text, so the benchmark discarded this measurement. 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 | 8,583.0 | — | 961 ms | — | — | 1 | — | — |
| Prefill | 16K | 1 | Valid | 6,700.0 | — | 2,440 ms | — | — | 1 | — | — |
| Prefill | 32K | 1 | Valid | 4,523.0 | — | 7,196 ms | — | — | 1 | — | — |
| Prefill | 64K | 1 | Valid | 2,711.0 | — | 23,950 ms | — | — | 1 | — | — |
| Sustained decode | 0 | 1 | Valid | 175.0 | 176.1 | 27 ms | 5.7 ms | 1.0 | 0 | 5,249 | 30.0 s |
| Sustained decode | 0 | 2 | Valid | 290.5 | 146.6 | 31 ms | 6.8 ms | 2.0 | 0 | 8,716 | 30.0 s |
| Sustained decode | 0 | 4 | Valid | 472.7 | 120.0 | 31 ms | 8.3 ms | 4.0 | 0 | 14,180 | 30.0 s |
| Sustained decode | 0 | 8 | Valid | 753.1 | 96.2 | 38 ms | 10.4 ms | 8.0 | 0 | 22,592 | 30.0 s |
| Sustained decode | 0 | 16 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 0 | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 0 | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 0 | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 1 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 16K | 2 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 16K | 4 | Over capacity | — | — | — | — | — | — | — | — |
| 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 | 130.1 | 130.3 | 61 ms | 7.7 ms | 1.0 | 0 | 3,902 | 30.0 s |
| Sustained decode | 32K | 2 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 4 | Over capacity | — | — | — | — | — | — | — | — |
| 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 | 104.5 | 104.6 | 113 ms | 9.6 ms | 1.0 | 0 | 3,135 | 30.0 s |
| Sustained decode | 64K | 2 | Over capacity | — | — | — | — | — | — | — | — |
| 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 | — | — | — | — | — | — | — | — |
Looped: the model kept repeating the same text, so the benchmark discarded this measurement. Over capacity: not run because this many requests at this context would exceed the server's KV cache.