Speed test matrix
NVIDIA-Nemotron-Nano-9B-v2-FP8-dynamic on RTX 3090 24GB
vllm · 0.32.0 · FP8 dynamic5 contexts (0–120K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 120K
Hardware
- Type
- Discrete GPU
- GPU
- RTX 3090
- VRAM
- 24 GB
- CPU
- Intel Core i7-9700K (8C)
- 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: none # 1x RTX 3090 24GB (Zagitta/gpu-unlock), 420 W limit, PCIe Gen3 x16 | Intel Core i7-9700K 8C | 32 GB DDR4 | Ubuntu 24.04.5, NVIDIA driver 580.178.04 vllm serve RedHatAI/NVIDIA-Nemotron-Nano-9B-v2-FP8-dynamic --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 context6,4431,302 ms TTFT
16K context6,2772,647 ms TTFT
32K context6,0175,497 ms TTFT
64K context5,55111,886 ms TTFT
120K context4,89625,239 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 | 85.433 ms TTFT | 164.646 ms TTFT | 308.289 ms TTFT | 548.3145 ms TTFT | 903.0259 ms TTFT | 1,330.2389 ms TTFT | 1,741.8656 ms TTFT | Over capacity |
| 16K | 83.2214 ms TTFT | 157.3282 ms TTFT | 292.2459 ms TTFT | 494.9792 ms TTFT | 750.31,530 ms TTFT | Over capacity | Over capacity | Over capacity |
| 32K | 81.2186 ms TTFT | 150.6186 ms TTFT | 278.4254 ms TTFT | 449.6427 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity |
| 64K | 77.4240 ms TTFT | 138.8292 ms TTFT | 254.2404 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity |
| 120K | 71.5326 ms TTFT | 122.1446 ms TTFT | 220.4678 ms TTFT | Over capacity | 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 | 6,443.0 | — | 1,302 ms | — | — | 1 | — | — |
| Prefill | 16K | 1 | Valid | 6,277.0 | — | 2,647 ms | — | — | 1 | — | — |
| Prefill | 32K | 1 | Valid | 6,017.0 | — | 5,497 ms | — | — | 1 | — | — |
| Prefill | 64K | 1 | Valid | 5,551.0 | — | 11,886 ms | — | — | 1 | — | — |
| Prefill | 120K | 1 | Valid | 4,896.0 | — | 25,239 ms | — | — | 1 | — | — |
| Sustained decode | 0 | 1 | Valid | 85.4 | 85.5 | 33 ms | 11.7 ms | 1.0 | 0 | 2,563 | 30.0 s |
| Sustained decode | 0 | 2 | Valid | 164.6 | 82.3 | 46 ms | 12.1 ms | 2.0 | 0 | 4,936 | 30.0 s |
| Sustained decode | 0 | 4 | Valid | 308.2 | 77.2 | 89 ms | 13.0 ms | 4.0 | 0 | 9,244 | 30.0 s |
| Sustained decode | 0 | 8 | Valid | 548.3 | 68.7 | 145 ms | 14.6 ms | 8.0 | 0 | 16,448 | 30.0 s |
| Sustained decode | 0 | 16 | Valid | 903.0 | 56.6 | 259 ms | 17.7 ms | 16.0 | 0 | 27,088 | 30.0 s |
| Sustained decode | 0 | 32 | Valid | 1,330.2 | 41.6 | 389 ms | 24.0 ms | 32.0 | 0 | 39,904 | 30.0 s |
| Sustained decode | 0 | 64 | Valid | 1,741.8 | 27.1 | 656 ms | 36.9 ms | 64.0 | 0 | 52,261 | 30.0 s |
| Sustained decode | 0 | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 1 | Valid | 83.2 | 83.3 | 214 ms | 12.0 ms | 1.0 | 0 | 2,495 | 30.0 s |
| Sustained decode | 16K | 2 | Valid | 157.3 | 78.5 | 282 ms | 12.7 ms | 2.0 | 0 | 4,718 | 30.0 s |
| Sustained decode | 16K | 4 | Valid | 292.2 | 72.7 | 459 ms | 13.8 ms | 4.0 | 0 | 8,763 | 30.0 s |
| Sustained decode | 16K | 8 | Valid | 494.9 | 60.9 | 792 ms | 16.4 ms | 8.0 | 0 | 14,840 | 30.0 s |
| Sustained decode | 16K | 16 | Valid | 750.3 | 45.3 | 1,530 ms | 22.1 ms | 16.0 | 0 | 22,512 | 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 | 81.2 | 81.3 | 186 ms | 12.3 ms | 1.0 | 0 | 2,436 | 30.0 s |
| Sustained decode | 32K | 2 | Valid | 150.6 | 75.3 | 186 ms | 13.3 ms | 2.0 | 0 | 4,518 | 30.0 s |
| Sustained decode | 32K | 4 | Valid | 278.4 | 69.5 | 254 ms | 14.4 ms | 4.0 | 0 | 8,348 | 30.0 s |
| Sustained decode | 32K | 8 | Valid | 449.6 | 56.1 | 427 ms | 17.8 ms | 8.0 | 0 | 13,488 | 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 | 77.4 | 77.4 | 240 ms | 12.9 ms | 1.0 | 0 | 2,320 | 30.0 s |
| Sustained decode | 64K | 2 | Valid | 138.8 | 69.4 | 292 ms | 14.4 ms | 2.0 | 0 | 4,165 | 30.0 s |
| Sustained decode | 64K | 4 | Valid | 254.2 | 63.3 | 404 ms | 15.8 ms | 4.0 | 0 | 7,624 | 30.0 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 | 120K | 1 | Valid | 71.5 | 71.6 | 326 ms | 14.0 ms | 1.0 | 0 | 2,144 | 30.0 s |
| Sustained decode | 120K | 2 | Valid | 122.1 | 61.1 | 446 ms | 16.4 ms | 2.0 | 0 | 3,662 | 30.0 s |
| Sustained decode | 120K | 4 | Valid | 220.4 | 54.9 | 678 ms | 18.2 ms | 4.0 | 0 | 6,608 | 30.0 s |
| Sustained decode | 120K | 8 | Over capacity | — | — | — | — | — | — | — | — |
| 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.