Speed test matrix
Qwen3.5-9B-FP8-dynamic on RTX 3090 24GB
vllm · 0.32.0 · FP8 dynamic2 contexts (0–8K) × 8 concurrency levels (1–128) · prefill at 8K
Hardware
- Type
- Discrete GPU
- GPU
- RTX 3090
- VRAM
- 24 GB
- CPU
- AMD Ryzen 9 7900X (12C)
- RAM
- 32 GB
- OS
- Ubuntu 24.04.5
Engine
- Engine
- vllm
- Version
- 0.32.0
- Quantization
- FP8 dynamic
- Backend
- cuda
- Repository
- github.com/vllm-project/vllm
- Build
- vllm/vllm-openai:v0.32.0-x86_64
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 RTX 3090 24GB (Zagitta/gpu-unlock), 350 W limit (VBIOS max), PCIe Gen4 x16 | AMD Ryzen 9 7900X 12C | 32 GB DDR5 | Ubuntu 24.04.5, NVIDIA driver 595.99.02
vllm serve RedHatAI/Qwen3.5-9B-FP8-dynamic --max-model-len 16384 --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 context6,0961,343 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 | 103.529 ms TTFT | 190.752 ms TTFT | 373.188 ms TTFT | 666.1135 ms TTFT | 1,134.8283 ms TTFT | Over capacity | Over capacity | Over capacity |
| 8K | 89.6114 ms TTFT | 176.2182 ms TTFT | 339.3403 ms TTFT | 540.6591 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 | 6,096.0 | — | 1,343 ms | — | — | 1 | — | — |
| Sustained decode | 0 | 1 | Valid | 103.5 | 103.9 | 29 ms | 9.6 ms | 1.0 | 0 | 3,103 | 30.0 s |
| Sustained decode | 0 | 2 | Valid | 190.7 | 97.4 | 52 ms | 10.3 ms | 2.0 | 0 | 5,718 | 30.0 s |
| Sustained decode | 0 | 4 | Valid | 373.1 | 95.1 | 88 ms | 10.5 ms | 4.0 | 0 | 11,184 | 30.0 s |
| Sustained decode | 0 | 8 | Valid | 666.1 | 85.2 | 135 ms | 11.7 ms | 8.0 | 0 | 19,970 | 30.0 s |
| Sustained decode | 0 | 16 | Valid | 1,134.8 | 72.2 | 283 ms | 13.9 ms | 16.0 | 0 | 34,021 | 30.0 s |
| Sustained decode | 0 | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 0 | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 0 | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 8K | 1 | Valid | 89.6 | 90.9 | 114 ms | 11.0 ms | 1.0 | 0 | 2,686 | 30.0 s |
| Sustained decode | 8K | 2 | Valid | 176.2 | 89.4 | 182 ms | 11.2 ms | 2.0 | 0 | 5,282 | 30.0 s |
| Sustained decode | 8K | 4 | Valid | 339.3 | 86.1 | 403 ms | 11.6 ms | 4.0 | 0 | 10,176 | 30.0 s |
| Sustained decode | 8K | 8 | Valid | 540.6 | 68.1 | 591 ms | 14.7 ms | 8.0 | 0 | 16,208 | 30.0 s |
| Sustained decode | 8K | 16 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 8K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 8K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 8K | 128 | Over capacity | — | — | — | — | — | — | — | — |
Over capacity: not run because this many requests at this context would exceed the server's KV cache.