Speed test matrix
Phi-4-reasoning-plus-NVFP4 on CMP 170HX 64GB
vllm · 0.32.0rc1.dev249+g18305948a · NVFP43 contexts (0–24K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 24K
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
- NVFP4
- 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 nvidia/Phi-4-reasoning-plus-NVFP4 --max-model-len 32768 --gpu-memory-utilization 0.94 --max-num-seqs 128 --enable-prefix-caching --trust-remote-code
Prefill
prompt tok/s by context length8K context3,2672,571 ms TTFT
16K context2,8845,691 ms TTFT
24K context2,6129,358 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 | 91.073 ms TTFT | 173.8128 ms TTFT | 328.3120 ms TTFT | 581.8155 ms TTFT | 939.6194 ms TTFT | 1,489.8287 ms TTFT | Over capacity | Over capacity |
| 16K | 76.8177 ms TTFT | 134.1206 ms TTFT | 210.0321 ms TTFT | 251.9419 ms TTFT | 341.3703 ms TTFT | Over capacity | Over capacity | Over capacity |
| 24K | 71.2236 ms TTFT | 120.5260 ms TTFT | 176.9403 ms TTFT | 193.1616 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 | 3,267.0 | — | 2,571 ms | — | — | 1 | — | — |
| Prefill | 16K | 1 | Valid | 2,884.0 | — | 5,691 ms | — | — | 1 | — | — |
| Prefill | 24K | 1 | Valid | 2,612.0 | — | 9,358 ms | — | — | 1 | — | — |
| Sustained decode | 0 | 1 | Valid | 91.0 | 91.3 | 73 ms | 11.0 ms | 1.0 | 0 | 2,729 | 30.0 s |
| Sustained decode | 0 | 2 | Valid | 173.8 | 87.4 | 128 ms | 11.4 ms | 2.0 | 0 | 5,212 | 30.0 s |
| Sustained decode | 0 | 4 | Valid | 328.3 | 83.1 | 120 ms | 12.0 ms | 4.0 | 0 | 9,848 | 30.0 s |
| Sustained decode | 0 | 8 | Valid | 581.8 | 74.5 | 155 ms | 13.4 ms | 8.0 | 0 | 17,456 | 30.0 s |
| Sustained decode | 0 | 16 | Valid | 939.6 | 60.4 | 194 ms | 16.6 ms | 16.0 | 0 | 28,208 | 30.0 s |
| Sustained decode | 0 | 32 | Valid | 1,489.8 | 48.0 | 287 ms | 20.8 ms | 32.0 | 0 | 44,704 | 30.0 s |
| Sustained decode | 0 | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 0 | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 1 | Valid | 76.8 | 77.1 | 177 ms | 13.0 ms | 1.0 | 0 | 2,305 | 30.0 s |
| Sustained decode | 16K | 2 | Valid | 134.1 | 67.2 | 206 ms | 14.9 ms | 2.0 | 0 | 4,022 | 30.0 s |
| Sustained decode | 16K | 4 | Valid | 210.0 | 52.8 | 321 ms | 19.0 ms | 4.0 | 0 | 6,300 | 30.0 s |
| Sustained decode | 16K | 8 | Valid | 251.9 | 31.6 | 419 ms | 31.6 ms | 8.0 | 0 | 7,552 | 30.0 s |
| Sustained decode | 16K | 16 | Valid | 341.3 | 21.4 | 703 ms | 46.6 ms | 16.0 | 0 | 10,240 | 30.0 s |
| Sustained decode | 16K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 24K | 1 | Valid | 71.2 | 71.5 | 236 ms | 14.0 ms | 1.0 | 0 | 2,134 | 30.0 s |
| Sustained decode | 24K | 2 | Valid | 120.5 | 60.5 | 260 ms | 16.5 ms | 2.0 | 0 | 3,616 | 30.0 s |
| Sustained decode | 24K | 4 | Valid | 176.9 | 44.4 | 403 ms | 22.5 ms | 4.0 | 0 | 5,304 | 30.0 s |
| Sustained decode | 24K | 8 | Valid | 193.1 | 24.3 | 616 ms | 41.2 ms | 8.0 | 0 | 5,792 | 30.0 s |
| Sustained decode | 24K | 16 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 24K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 24K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 24K | 128 | Over capacity | — | — | — | — | — | — | — | — |
Over capacity: not run because this many requests at this context would exceed the server's KV cache.