Speed test matrix
Qwen3-8B-AWQ on CMP 170HX 64GB
vllm · 0.32.0rc1.dev249+g18305948a · AWQ3 contexts (0–32K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K
Hardware
- Type
- Discrete GPU
- GPU
- CMP 170HX
- VRAM
- 64 GB
- CPU
- 2x Intel Xeon E5-2630 v3
- RAM
- 64 GB
- OS
- Ubuntu 26.04.1
Engine
- Engine
- vllm
- Version
- 0.32.0rc1.dev249+g18305948a
- Quantization
- AWQ
- Backend
- cuda
- Repository
- github.com/vllm-project/vllm
- Build
- ghcr.io/morrowmake/vllm-cmp170hx@sha256:5b45c4d6feafefe2eba962375b0a9a0b64fd22b9e05abb4d1966f14a32a2db79
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 NVIDIA CMP 170HX 64GB (cmpunlocker), 180 W limit, PCIe Gen2 x4 | 2x Intel Xeon E5-2630 v3 | 64 GB DDR4 | Ubuntu 26.04.1, NVIDIA driver 610.57.04 | vLLM: Morrowmake vllm-cmp170hx image 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
Prefill
prompt tok/s by context length8K context6,0831,349 ms TTFT
16K context5,3273,049 ms TTFT
32K context4,3737,395 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 | 142.576 ms TTFT | 263.8102 ms TTFT | 484.6113 ms TTFT | 846.7136 ms TTFT | 1,343.1175 ms TTFT | 1,989.9299 ms TTFT | Over capacity | Over capacity |
| 16K | 110.9186 ms TTFT | 182.4213 ms TTFT | 279.4278 ms TTFT | Looped | Over capacity | Over capacity | Over capacity | Over capacity |
| 32K | 90.2255 ms TTFT | 135.6363 ms TTFT | 186.5533 ms TTFT | 232.1903 ms TTFT | 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 | 6,083.0 | — | 1,349 ms | — | — | 1 | — | — |
| Prefill | 16K | 1 | Valid | 5,327.0 | — | 3,049 ms | — | — | 1 | — | — |
| Prefill | 32K | 1 | Valid | 4,373.0 | — | 7,395 ms | — | — | 1 | — | — |
| Sustained decode | 0 | 1 | Valid | 142.5 | 144.0 | 76 ms | 6.9 ms | 1.0 | 0 | 4,276 | 30.0 s |
| Sustained decode | 0 | 2 | Valid | 263.8 | 134.1 | 102 ms | 7.5 ms | 2.0 | 0 | 7,914 | 30.0 s |
| Sustained decode | 0 | 4 | Valid | 484.6 | 124.5 | 113 ms | 8.0 ms | 4.0 | 0 | 14,536 | 30.0 s |
| Sustained decode | 0 | 8 | Valid | 846.7 | 110.4 | 136 ms | 9.1 ms | 8.0 | 0 | 25,406 | 30.0 s |
| Sustained decode | 0 | 16 | Valid | 1,343.1 | 89.3 | 175 ms | 11.2 ms | 16.0 | 0 | 40,288 | 30.0 s |
| Sustained decode | 0 | 32 | Valid | 1,989.9 | 66.8 | 299 ms | 15.0 ms | 32.0 | 0 | 59,689 | 30.0 s |
| Sustained decode | 0 | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 0 | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 1 | Valid | 110.9 | 111.3 | 186 ms | 9.0 ms | 1.0 | 0 | 3,326 | 30.0 s |
| Sustained decode | 16K | 2 | Valid | 182.4 | 91.7 | 213 ms | 10.9 ms | 2.0 | 0 | 5,470 | 30.0 s |
| Sustained decode | 16K | 4 | Valid | 279.4 | 70.3 | 278 ms | 14.2 ms | 4.0 | 0 | 8,380 | 30.0 s |
| Sustained decode | 16K | 8 | Looped | — | — | — | — | — | 0 | 0 | — |
| 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 | 90.2 | 90.6 | 255 ms | 11.0 ms | 1.0 | 0 | 2,706 | 30.0 s |
| Sustained decode | 32K | 2 | Valid | 135.6 | 68.1 | 363 ms | 14.7 ms | 2.0 | 0 | 4,068 | 30.0 s |
| Sustained decode | 32K | 4 | Valid | 186.5 | 46.8 | 533 ms | 21.4 ms | 4.0 | 0 | 5,593 | 30.0 s |
| Sustained decode | 32K | 8 | Valid | 232.1 | 29.0 | 903 ms | 34.5 ms | 8.0 | 0 | 6,960 | 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 | — | — | — | — | — | — | — | — |
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.