Speed test matrix
Qwen3-8B-FP8 on CMP 170HX 64GB
vllm · 0.32.0rc1.dev249+g18305948a · FP83 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
- FP8
- 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-FP8 --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 context5,5671,474 ms TTFT
16K context4,9263,298 ms TTFT
32K context4,1127,864 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 | 107.856 ms TTFT | 205.499 ms TTFT | 390.5107 ms TTFT | 711.8136 ms TTFT | 1,225.2201 ms TTFT | 1,889.2287 ms TTFT | Over capacity | Over capacity |
| 16K | 88.9171 ms TTFT | 154.8222 ms TTFT | 246.6284 ms TTFT | Looped | Over capacity | Over capacity | Over capacity | Over capacity |
| 32K | 75.5247 ms TTFT | Looped | Looped | 221.2770 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 | 5,567.0 | — | 1,474 ms | — | — | 1 | — | — |
| Prefill | 16K | 1 | Valid | 4,926.0 | — | 3,298 ms | — | — | 1 | — | — |
| Prefill | 32K | 1 | Valid | 4,112.0 | — | 7,864 ms | — | — | 1 | — | — |
| Sustained decode | 0 | 1 | Valid | 107.8 | 108.5 | 56 ms | 9.2 ms | 1.0 | 0 | 3,232 | 30.0 s |
| Sustained decode | 0 | 2 | Valid | 205.4 | 103.8 | 99 ms | 9.6 ms | 2.0 | 0 | 6,162 | 30.0 s |
| Sustained decode | 0 | 4 | Valid | 390.5 | 99.3 | 107 ms | 10.1 ms | 4.0 | 0 | 11,712 | 30.0 s |
| Sustained decode | 0 | 8 | Valid | 711.8 | 91.4 | 136 ms | 10.9 ms | 8.0 | 0 | 21,368 | 30.0 s |
| Sustained decode | 0 | 16 | Valid | 1,225.2 | 80.2 | 201 ms | 12.5 ms | 16.0 | 0 | 36,752 | 30.0 s |
| Sustained decode | 0 | 32 | Valid | 1,889.2 | 62.6 | 287 ms | 16.0 ms | 32.0 | 0 | 56,672 | 30.0 s |
| Sustained decode | 0 | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 0 | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 1 | Valid | 88.9 | 89.2 | 171 ms | 11.2 ms | 1.0 | 0 | 2,667 | 30.0 s |
| Sustained decode | 16K | 2 | Valid | 154.8 | 77.7 | 222 ms | 12.9 ms | 2.0 | 0 | 4,642 | 30.0 s |
| Sustained decode | 16K | 4 | Valid | 246.6 | 61.9 | 284 ms | 16.1 ms | 4.0 | 0 | 7,396 | 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 | 75.5 | 75.8 | 247 ms | 13.2 ms | 1.0 | 0 | 2,266 | 30.0 s |
| Sustained decode | 32K | 2 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 32K | 4 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 32K | 8 | Valid | 221.2 | 27.8 | 770 ms | 36.0 ms | 8.0 | 0 | 6,632 | 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.