Speed test matrix
Llama-3.2-3B-Instruct-FP8-dynamic on RTX 3060 12GB
vllm · 0.32.0 · FP8 dynamic4 contexts (0–48K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 48K
Hardware
- Type
- Discrete GPU
- GPU
- RTX 3060
- VRAM
- 12 GB
- CPU
- Intel Xeon E5-2680 v4 (14C)
- 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 3060 12GB, stock 170 W limit, PCIe Gen3 x16 | Intel Xeon E5-2680 v4 14C | 32 GB DDR4-2400 | Ubuntu 24.04.5, NVIDIA driver 580.178.04 vllm serve RedHatAI/Llama-3.2-3B-Instruct-FP8-dynamic --max-model-len 57344 --gpu-memory-utilization 0.94 --max-num-seqs 128 --enable-prefix-caching --trust-remote-code
Prefill
prompt tok/s by context length8K context3,2652,515 ms TTFT
16K context2,6806,056 ms TTFT
32K context2,00516,098 ms TTFT
48K context1,59330,356 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 | 77.849 ms TTFT | 146.673 ms TTFT | 267.284 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity |
| 16K | 51.8156 ms TTFT | 77.2189 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity |
| 32K | 38.7239 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity |
| 48K | 30.7362 ms TTFT | Over capacity | Over capacity | 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 | 3,265.0 | — | 2,515 ms | — | — | 1 | — | — |
| Prefill | 16K | 1 | Valid | 2,680.0 | — | 6,056 ms | — | — | 1 | — | — |
| Prefill | 32K | 1 | Valid | 2,005.0 | — | 16,098 ms | — | — | 1 | — | — |
| Prefill | 48K | 1 | Valid | 1,593.0 | — | 30,356 ms | — | — | 1 | — | — |
| Sustained decode | 0 | 1 | Valid | 77.8 | 78.6 | 49 ms | 12.7 ms | 1.0 | 0 | 2,333 | 30.0 s |
| Sustained decode | 0 | 2 | Valid | 146.6 | 74.3 | 73 ms | 13.5 ms | 2.0 | 0 | 4,397 | 30.0 s |
| Sustained decode | 0 | 4 | Valid | 267.2 | 68.5 | 84 ms | 14.6 ms | 4.0 | 0 | 8,012 | 30.0 s |
| Sustained decode | 0 | 8 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 0 | 16 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 0 | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 0 | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 0 | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 1 | Valid | 51.8 | 52.0 | 156 ms | 19.2 ms | 1.0 | 0 | 1,554 | 30.0 s |
| Sustained decode | 16K | 2 | Valid | 77.2 | 38.7 | 189 ms | 25.8 ms | 2.0 | 0 | 2,316 | 30.0 s |
| Sustained decode | 16K | 4 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 8 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 16 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 1 | Valid | 38.7 | 38.7 | 239 ms | 25.8 ms | 1.0 | 0 | 1,159 | 30.0 s |
| Sustained decode | 32K | 2 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 4 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 8 | Over capacity | — | — | — | — | — | — | — | — |
| 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 | 48K | 1 | Valid | 30.7 | 30.8 | 362 ms | 32.5 ms | 1.0 | 0 | 921 | 30.0 s |
| Sustained decode | 48K | 2 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 48K | 4 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 48K | 8 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 48K | 16 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 48K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 48K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 48K | 128 | Over capacity | — | — | — | — | — | — | — | — |
Over capacity: not run because this many requests at this context would exceed the server's KV cache.