Speed test matrix
Qwen3.5-27B-AWQ-4bit on CMP 170HX 64GB
vllm · 0.32.0rc1.dev249+g18305948a · AWQ 4-bit6 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
- AWQ 4-bit
- 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 cyankiwi/Qwen3.5-27B-AWQ-4bit --max-model-len 262144 --gpu-memory-utilization 0.94 --max-num-seqs 128 --enable-prefix-caching --trust-remote-code --language-model-only
Prefill
prompt tok/s by context length8K context1,9764,122 ms TTFT
16K context1,8518,707 ms TTFT
32K context1,73718,466 ms TTFT
64K context1,53841,619 ms TTFT
128K context1,245102,738 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 | 52.7168 ms TTFT | 101.8300 ms TTFT | 194.2310 ms TTFT | 363.3488 ms TTFT | 602.4854 ms TTFT | 902.51,462 ms TTFT | 1,140.72,672 ms TTFT | Over capacity |
| 16K | 49.8394 ms TTFT | 92.8576 ms TTFT | 166.8854 ms TTFT | 281.21,472 ms TTFT | 409.82,767 ms TTFT | Over capacity | Over capacity | Over capacity |
| 32K | 47.5673 ms TTFT | 85.6964 ms TTFT | 146.01,472 ms TTFT | 229.32,599 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity |
| 64K | 43.7846 ms TTFT | 74.01,095 ms TTFT | 117.01,680 ms TTFT | 167.62,791 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity |
| 128K | 37.51,018 ms TTFT | 58.31,227 ms TTFT | 83.61,754 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity |
| 248K | 27.8Requests queued2,648 ms TTFT | 37.0Requests queued3,342 ms TTFT | 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 | 1,976.0 | — | 4,122 ms | — | — | 1 | — | — |
| Prefill | 16K | 1 | Valid | 1,851.0 | — | 8,707 ms | — | — | 1 | — | — |
| Prefill | 32K | 1 | Valid | 1,737.0 | — | 18,466 ms | — | — | 1 | — | — |
| Prefill | 64K | 1 | Valid | 1,538.0 | — | 41,619 ms | — | — | 1 | — | — |
| Prefill | 128K | 1 | Valid | 1,245.0 | — | 102,738 ms | — | — | 1 | — | — |
| Sustained decode | 0 | 1 | Valid | 52.7 | 52.7 | 168 ms | 19.0 ms | 1.0 | 0 | 1,579 | 30.0 s |
| Sustained decode | 0 | 2 | Valid | 101.8 | 51.0 | 300 ms | 19.6 ms | 2.0 | 0 | 3,052 | 30.0 s |
| Sustained decode | 0 | 4 | Valid | 194.2 | 48.7 | 310 ms | 20.5 ms | 4.0 | 0 | 5,827 | 30.0 s |
| Sustained decode | 0 | 8 | Valid | 363.3 | 45.6 | 488 ms | 21.9 ms | 8.0 | 0 | 10,896 | 30.0 s |
| Sustained decode | 0 | 16 | Valid | 602.4 | 37.8 | 854 ms | 26.4 ms | 16.0 | 0 | 18,064 | 30.0 s |
| Sustained decode | 0 | 32 | Valid | 902.5 | 28.3 | 1,462 ms | 35.4 ms | 32.0 | 0 | 27,104 | 30.0 s |
| Sustained decode | 0 | 64 | Valid | 1,140.7 | 17.8 | 2,672 ms | 56.3 ms | 64.0 | 0 | 34,240 | 30.0 s |
| Sustained decode | 0 | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 1 | Valid | 49.8 | 49.8 | 394 ms | 20.1 ms | 1.0 | 0 | 1,492 | 30.0 s |
| Sustained decode | 16K | 2 | Valid | 92.8 | 46.3 | 576 ms | 21.6 ms | 2.0 | 0 | 2,784 | 30.0 s |
| Sustained decode | 16K | 4 | Valid | 166.8 | 41.3 | 854 ms | 24.2 ms | 4.0 | 0 | 5,000 | 30.0 s |
| Sustained decode | 16K | 8 | Valid | 281.2 | 34.3 | 1,472 ms | 29.2 ms | 8.0 | 0 | 8,432 | 30.0 s |
| Sustained decode | 16K | 16 | Valid | 409.8 | 24.1 | 2,767 ms | 41.4 ms | 16.0 | 0 | 12,288 | 30.0 s |
| Sustained decode | 16K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 1 | Valid | 47.5 | 47.6 | 673 ms | 21.0 ms | 1.0 | 0 | 1,425 | 30.0 s |
| Sustained decode | 32K | 2 | Valid | 85.6 | 42.6 | 964 ms | 23.5 ms | 2.0 | 0 | 2,568 | 30.0 s |
| Sustained decode | 32K | 4 | Valid | 146.0 | 35.8 | 1,472 ms | 27.9 ms | 4.0 | 0 | 4,376 | 30.0 s |
| Sustained decode | 32K | 8 | Valid | 229.3 | 27.3 | 2,599 ms | 36.7 ms | 8.0 | 0 | 6,872 | 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 | 43.7 | 43.8 | 846 ms | 22.8 ms | 1.0 | 0 | 1,310 | 30.0 s |
| Sustained decode | 64K | 2 | Valid | 74.0 | 36.8 | 1,095 ms | 27.2 ms | 2.0 | 0 | 2,218 | 30.0 s |
| Sustained decode | 64K | 4 | Valid | 117.0 | 28.7 | 1,680 ms | 34.8 ms | 4.0 | 0 | 3,508 | 30.0 s |
| Sustained decode | 64K | 8 | Valid | 167.6 | 20.0 | 2,791 ms | 50.1 ms | 8.0 | 0 | 5,024 | 30.0 s |
| 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 | 37.5 | 37.7 | 1,018 ms | 26.5 ms | 1.0 | 0 | 1,124 | 30.0 s |
| Sustained decode | 128K | 2 | Valid | 58.3 | 29.2 | 1,227 ms | 34.3 ms | 2.0 | 0 | 1,746 | 30.0 s |
| Sustained decode | 128K | 4 | Valid | 83.6 | 20.8 | 1,754 ms | 48.1 ms | 4.0 | 0 | 2,508 | 30.0 s |
| 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 | 27.8 | 29.9 | 2,648 ms | 33.5 ms | 0.9 | 0 | 832 | 30.0 s |
| Sustained decode | 248K | 2 | Requests queued | 37.0 | 20.4 | 3,342 ms | 48.9 ms | 1.8 | 0 | 1,110 | 30.0 s |
| 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.