Speed test matrix
Qwen3.5-27B-FP8 on CMP 170HX 64GB
vllm · 0.32.0rc1.dev249+g18305948a · FP86 contexts (0–248K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 128K
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: MTP, 2 tokens lookahead
# 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/Qwen3.5-27B-FP8 --max-model-len 262144 --gpu-memory-utilization 0.94 --max-num-seqs 128 --enable-prefix-caching --trust-remote-code --language-model-only --speculative-config {"method":"mtp","num_speculative_tokens":2}Prefill
prompt tok/s by context length8K context1,6644,895 ms TTFT
16K context1,60310,056 ms TTFT
32K context1,53320,931 ms TTFT
64K context1,36746,825 ms TTFT
128K context1,123113,871 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 | 72.8209 ms TTFT | 140.4382 ms TTFT | 256.5426 ms TTFT | 451.5620 ms TTFT | 697.01,037 ms TTFT | 769.81,658 ms TTFT | Over capacity | Over capacity |
| 16K | 54.4754 ms TTFT | 104.71,218 ms TTFT | 154.12,369 ms TTFT | 238.93,999 ms TTFT | 298.66,557 ms TTFT | Over capacity | Over capacity | Over capacity |
| 32K | 43.6838 ms TTFT | 83.11,384 ms TTFT | 112.32,666 ms TTFT | 161.14,438 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity |
| 64K | 30.71,172 ms TTFT | 58.01,748 ms TTFT | 72.13,185 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity |
| 128K | 18.92,483 ms TTFT | 37.84,090 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity |
| 248K | 10.8Requests queued3,790 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity |
Requests queued: the server could not run all requests at once, so some waited in a queue. 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 | 1,664.0 | — | 4,895 ms | — | — | 1 | — | — |
| Prefill | 16K | 1 | Valid | 1,603.0 | — | 10,056 ms | — | — | 1 | — | — |
| Prefill | 32K | 1 | Valid | 1,533.0 | — | 20,931 ms | — | — | 1 | — | — |
| Prefill | 64K | 1 | Valid | 1,367.0 | — | 46,825 ms | — | — | 1 | — | — |
| Prefill | 128K | 1 | Valid | 1,123.0 | — | 113,871 ms | — | — | 1 | — | — |
| Sustained decode | 0 | 1 | Valid | 72.8 | 73.7 | 209 ms | 13.6 ms | 1.0 | 0 | 2,181 | 30.0 s |
| Sustained decode | 0 | 2 | Valid | 140.4 | 70.8 | 382 ms | 14.1 ms | 2.0 | 0 | 4,207 | 30.0 s |
| Sustained decode | 0 | 4 | Valid | 256.5 | 65.1 | 426 ms | 15.4 ms | 4.0 | 0 | 7,686 | 30.0 s |
| Sustained decode | 0 | 8 | Valid | 451.5 | 57.1 | 620 ms | 17.5 ms | 8.0 | 0 | 13,534 | 30.0 s |
| Sustained decode | 0 | 16 | Valid | 697.0 | 44.9 | 1,037 ms | 22.3 ms | 16.0 | 0 | 20,912 | 30.0 s |
| Sustained decode | 0 | 32 | Valid | 769.8 | 24.6 | 1,658 ms | 40.7 ms | 32.0 | 0 | 23,073 | 30.0 s |
| Sustained decode | 0 | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 0 | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 1 | Valid | 54.4 | 55.1 | 754 ms | 18.1 ms | 1.0 | 0 | 1,632 | 30.0 s |
| Sustained decode | 16K | 2 | Valid | 104.7 | 52.7 | 1,218 ms | 19.0 ms | 2.0 | 0 | 3,135 | 30.0 s |
| Sustained decode | 16K | 4 | Valid | 154.1 | 38.5 | 2,369 ms | 26.0 ms | 4.0 | 0 | 4,617 | 30.0 s |
| Sustained decode | 16K | 8 | Valid | 238.9 | 29.3 | 3,999 ms | 34.1 ms | 8.0 | 0 | 7,160 | 30.0 s |
| Sustained decode | 16K | 16 | Valid | 298.6 | 17.4 | 6,557 ms | 57.4 ms | 16.0 | 0 | 8,927 | 29.9 s |
| Sustained decode | 16K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 1 | Valid | 43.6 | 44.0 | 838 ms | 22.7 ms | 1.0 | 0 | 1,308 | 30.0 s |
| Sustained decode | 32K | 2 | Valid | 83.1 | 41.9 | 1,384 ms | 23.9 ms | 2.0 | 0 | 2,492 | 30.0 s |
| Sustained decode | 32K | 4 | Valid | 112.3 | 28.5 | 2,666 ms | 35.1 ms | 4.0 | 0 | 3,365 | 30.0 s |
| Sustained decode | 32K | 8 | Valid | 161.1 | 19.8 | 4,438 ms | 50.5 ms | 8.0 | 0 | 4,826 | 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 | — | — | — | — | — | — | — | — |
| Sustained decode | 64K | 1 | Valid | 30.7 | 31.1 | 1,172 ms | 32.2 ms | 1.0 | 0 | 918 | 29.9 s |
| Sustained decode | 64K | 2 | Valid | 58.0 | 29.4 | 1,748 ms | 34.0 ms | 2.0 | 0 | 1,740 | 30.0 s |
| Sustained decode | 64K | 4 | Valid | 72.1 | 18.0 | 3,185 ms | 55.5 ms | 4.0 | 0 | 2,156 | 29.9 s |
| 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 | — | — | — | — | — | — | — | — |
| Sustained decode | 128K | 1 | Valid | 18.9 | 19.2 | 2,483 ms | 52.1 ms | 1.0 | 0 | 567 | 30.0 s |
| Sustained decode | 128K | 2 | Valid | 37.8 | 19.0 | 4,090 ms | 52.8 ms | 2.0 | 0 | 1,134 | 30.0 s |
| Sustained decode | 128K | 4 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 128K | 8 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 128K | 16 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 128K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 128K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 128K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 248K | 1 | Requests queued | 10.8 | 12.0 | 3,790 ms | 83.1 ms | 0.9 | 0 | 322 | 29.9 s |
| Sustained decode | 248K | 2 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 248K | 4 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 248K | 8 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 248K | 16 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 248K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 248K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 248K | 128 | Over capacity | — | — | — | — | — | — | — | — |
Requests queued: the server could not run all requests at once, so some waited in a queue. Over capacity: not run because this many requests at this context would exceed the server's KV cache.