Speed test matrix
gpt-oss-20b on CMP 170HX 64GB
vllm · 0.32.0rc1.dev249+g18305948a · MXFP45 contexts (0–120K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 120K
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
- MXFP4
- 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 openai/gpt-oss-20b --max-model-len 131072 --gpu-memory-utilization 0.94 --max-num-seqs 128 --enable-prefix-caching --trust-remote-code
Prefill
prompt tok/s by context length8K context8,773924 ms TTFT
16K context6,8882,322 ms TTFT
32K context4,9056,481 ms TTFT
64K context3,13020,248 ms TTFT
120K context1,88862,846 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 | 214.171 ms TTFT | 307.8130 ms TTFT | 438.0180 ms TTFT | 740.8190 ms TTFT | 1,116.3251 ms TTFT | 1,728.8318 ms TTFT | 2,827.4594 ms TTFT | 3,816.0868 ms TTFT |
| 16K | 161.2147 ms TTFT | 251.0227 ms TTFT | 377.4359 ms TTFT | 486.4608 ms TTFT | 629.71,072 ms TTFT | 1,027.41,821 ms TTFT | 1,525.22,969 ms TTFT | Over capacity |
| 32K | 147.7233 ms TTFT | 232.2351 ms TTFT | 252.1588 ms TTFT | 437.8980 ms TTFT | 523.41,846 ms TTFT | 599.33,203 ms TTFT | Over capacity | Over capacity |
| 64K | 106.7335 ms TTFT | 159.7538 ms TTFT | 227.5984 ms TTFT | 274.91,636 ms TTFT | 342.62,850 ms TTFT | Over capacity | Over capacity | Over capacity |
| 120K | 76.6531 ms TTFT | 117.9770 ms TTFT | 154.81,362 ms TTFT | 100.62,419 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 | 8,773.0 | — | 924 ms | — | — | 1 | — | — |
| Prefill | 16K | 1 | Valid | 6,888.0 | — | 2,322 ms | — | — | 1 | — | — |
| Prefill | 32K | 1 | Valid | 4,905.0 | — | 6,481 ms | — | — | 1 | — | — |
| Prefill | 64K | 1 | Valid | 3,130.0 | — | 20,248 ms | — | — | 1 | — | — |
| Prefill | 120K | 1 | Valid | 1,888.0 | — | 62,846 ms | — | — | 1 | — | — |
| Sustained decode | 0 | 1 | Valid | 214.1 | 194.9 | 71 ms | 5.1 ms | 1.0 | 0 | 6,423 | 30.0 s |
| Sustained decode | 0 | 2 | Valid | 307.8 | 152.7 | 130 ms | 6.5 ms | 2.0 | 0 | 9,229 | 30.0 s |
| Sustained decode | 0 | 4 | Valid | 438.0 | 120.8 | 180 ms | 8.3 ms | 4.0 | 0 | 13,140 | 30.0 s |
| Sustained decode | 0 | 8 | Valid | 740.8 | 93.5 | 190 ms | 10.7 ms | 8.0 | 0 | 22,224 | 30.0 s |
| Sustained decode | 0 | 16 | Valid | 1,116.3 | 76.4 | 251 ms | 13.1 ms | 16.0 | 0 | 33,477 | 30.0 s |
| Sustained decode | 0 | 32 | Valid | 1,728.8 | 62.9 | 318 ms | 15.9 ms | 32.0 | 0 | 51,855 | 30.0 s |
| Sustained decode | 0 | 64 | Valid | 2,827.4 | 50.9 | 594 ms | 19.6 ms | 64.0 | 0 | 84,824 | 30.0 s |
| Sustained decode | 0 | 128 | Valid | 3,816.0 | 36.4 | 868 ms | 27.5 ms | 128.0 | 0 | 114,531 | 30.0 s |
| Sustained decode | 16K | 1 | Valid | 161.2 | 162.0 | 147 ms | 6.2 ms | 1.0 | 0 | 4,816 | 29.9 s |
| Sustained decode | 16K | 2 | Valid | 251.0 | 126.6 | 227 ms | 7.9 ms | 2.0 | 0 | 7,528 | 30.0 s |
| Sustained decode | 16K | 4 | Valid | 377.4 | 94.2 | 359 ms | 10.6 ms | 4.0 | 0 | 11,325 | 30.0 s |
| Sustained decode | 16K | 8 | Valid | 486.4 | 70.5 | 608 ms | 14.2 ms | 8.0 | 0 | 14,588 | 30.0 s |
| Sustained decode | 16K | 16 | Valid | 629.7 | 51.5 | 1,072 ms | 19.4 ms | 16.0 | 0 | 18,910 | 30.0 s |
| Sustained decode | 16K | 32 | Valid | 1,027.4 | 40.1 | 1,821 ms | 25.0 ms | 32.0 | 0 | 30,807 | 30.0 s |
| Sustained decode | 16K | 64 | Valid | 1,525.2 | 29.5 | 2,969 ms | 33.9 ms | 64.0 | 0 | 45,722 | 30.0 s |
| Sustained decode | 16K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 1 | Valid | 147.7 | 138.4 | 233 ms | 7.2 ms | 1.0 | 0 | 4,408 | 29.9 s |
| Sustained decode | 32K | 2 | Valid | 232.2 | 110.9 | 351 ms | 9.0 ms | 2.0 | 0 | 6,965 | 30.0 s |
| Sustained decode | 32K | 4 | Valid | 252.1 | 77.7 | 588 ms | 12.9 ms | 4.0 | 0 | 7,563 | 30.0 s |
| Sustained decode | 32K | 8 | Valid | 437.8 | 56.2 | 980 ms | 17.8 ms | 8.0 | 0 | 13,130 | 30.0 s |
| Sustained decode | 32K | 16 | Valid | 523.4 | 38.7 | 1,846 ms | 25.9 ms | 16.0 | 0 | 15,694 | 30.0 s |
| Sustained decode | 32K | 32 | Valid | 599.3 | 29.0 | 3,203 ms | 34.5 ms | 32.0 | 0 | 17,965 | 30.0 s |
| Sustained decode | 32K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 64K | 1 | Valid | 106.7 | 107.1 | 335 ms | 9.3 ms | 1.0 | 0 | 3,204 | 30.0 s |
| Sustained decode | 64K | 2 | Valid | 159.7 | 86.8 | 538 ms | 11.5 ms | 2.0 | 0 | 4,786 | 30.0 s |
| Sustained decode | 64K | 4 | Valid | 227.5 | 57.7 | 984 ms | 17.3 ms | 4.0 | 0 | 6,822 | 30.0 s |
| Sustained decode | 64K | 8 | Valid | 274.9 | 40.2 | 1,636 ms | 24.8 ms | 8.0 | 0 | 8,247 | 30.0 s |
| Sustained decode | 64K | 16 | Valid | 342.6 | 26.2 | 2,850 ms | 38.2 ms | 16.0 | 0 | 10,271 | 30.0 s |
| Sustained decode | 64K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 64K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 64K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 120K | 1 | Valid | 76.6 | 77.0 | 531 ms | 13.0 ms | 1.0 | 0 | 2,297 | 30.0 s |
| Sustained decode | 120K | 2 | Valid | 117.9 | 63.3 | 770 ms | 15.8 ms | 2.0 | 0 | 3,531 | 29.9 s |
| Sustained decode | 120K | 4 | Valid | 154.8 | 39.3 | 1,362 ms | 25.4 ms | 4.0 | 0 | 4,644 | 30.0 s |
| Sustained decode | 120K | 8 | Valid | 100.6 | 27.2 | 2,419 ms | 36.7 ms | 8.0 | 0 | 3,015 | 30.0 s |
| Sustained decode | 120K | 16 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 120K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 120K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 120K | 128 | Over capacity | — | — | — | — | — | — | — | — |
Over capacity: not run because this many requests at this context would exceed the server's KV cache.