Speed test matrix
Fara1.5-4B-AutoRound-W4A16-RTN on RTX 3090 24GB
vllm · 0.32.0 · AutoRound W4A166 contexts (0–248K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 128K
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
- AutoRound W4A16
- 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 LeaderboardModel1/Fara1.5-4B-AutoRound-W4A16-RTN --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 context11.8K691 ms TTFT
16K context11K1,474 ms TTFT
32K context9,9333,251 ms TTFT
64K context8,2747,786 ms TTFT
128K context6,19520,772 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 | Looped | 345.437 ms TTFT | 636.766 ms TTFT | 1,084.593 ms TTFT | 1,740.8174 ms TTFT | 2,576.5244 ms TTFT | 3,297.1460 ms TTFT | Over capacity |
| 16K | 163.991 ms TTFT | 293.4130 ms TTFT | 531.4184 ms TTFT | 867.7305 ms TTFT | 1,211.8578 ms TTFT | Over capacity | Over capacity | Over capacity |
| 32K | 149.6106 ms TTFT | 253.6141 ms TTFT | 454.3210 ms TTFT | 690.5345 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity |
| 64K | Looped | Looped | 343.8350 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity |
| 128K | Looped | 138.5600 ms TTFT | 228.6834 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity |
| 248K | 68.1915 ms TTFT | 88.21,169 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity |
Looped: the model kept repeating the same text, so the benchmark discarded this measurement. 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 | 11.8K | — | 691 ms | — | — | 1 | — | — |
| Prefill | 16K | 1 | Valid | 11K | — | 1,474 ms | — | — | 1 | — | — |
| Prefill | 32K | 1 | Valid | 9,933.0 | — | 3,251 ms | — | — | 1 | — | — |
| Prefill | 64K | 1 | Valid | 8,274.0 | — | 7,786 ms | — | — | 1 | — | — |
| Prefill | 128K | 1 | Valid | 6,195.0 | — | 20,772 ms | — | — | 1 | — | — |
| Sustained decode | 0 | 1 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 0 | 2 | Valid | 345.4 | 173.9 | 37 ms | 5.7 ms | 2.0 | 0 | 10,362 | 30.0 s |
| Sustained decode | 0 | 4 | Valid | 636.7 | 160.8 | 66 ms | 6.2 ms | 4.0 | 0 | 19,100 | 30.0 s |
| Sustained decode | 0 | 8 | Valid | 1,084.5 | 137.9 | 93 ms | 7.3 ms | 8.0 | 0 | 32,536 | 30.0 s |
| Sustained decode | 0 | 16 | Valid | 1,740.8 | 111.2 | 174 ms | 9.0 ms | 16.0 | 0 | 52,224 | 30.0 s |
| Sustained decode | 0 | 32 | Valid | 2,576.5 | 82.2 | 244 ms | 12.2 ms | 32.0 | 0 | 77,303 | 30.0 s |
| Sustained decode | 0 | 64 | Valid | 3,297.1 | 52.5 | 460 ms | 19.0 ms | 64.0 | 0 | 98,880 | 30.0 s |
| Sustained decode | 0 | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 1 | Valid | 163.9 | 164.2 | 91 ms | 6.1 ms | 1.0 | 0 | 4,916 | 30.0 s |
| Sustained decode | 16K | 2 | Valid | 293.4 | 147.0 | 130 ms | 6.8 ms | 2.0 | 0 | 8,802 | 30.0 s |
| Sustained decode | 16K | 4 | Valid | 531.4 | 133.2 | 184 ms | 7.5 ms | 4.0 | 0 | 15,940 | 30.0 s |
| Sustained decode | 16K | 8 | Valid | 867.7 | 108.4 | 305 ms | 9.2 ms | 8.0 | 0 | 26,024 | 30.0 s |
| Sustained decode | 16K | 16 | Valid | 1,211.8 | 75.1 | 578 ms | 13.3 ms | 16.0 | 0 | 36,355 | 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 | 149.6 | 149.9 | 106 ms | 6.7 ms | 1.0 | 0 | 4,487 | 30.0 s |
| Sustained decode | 32K | 2 | Valid | 253.6 | 127.1 | 141 ms | 7.9 ms | 2.0 | 0 | 7,606 | 30.0 s |
| Sustained decode | 32K | 4 | Valid | 454.3 | 113.9 | 210 ms | 8.8 ms | 4.0 | 0 | 13,628 | 30.0 s |
| Sustained decode | 32K | 8 | Valid | 690.5 | 86.3 | 345 ms | 11.6 ms | 8.0 | 0 | 20,712 | 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 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 64K | 2 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 64K | 4 | Valid | 343.8 | 86.1 | 350 ms | 11.6 ms | 4.0 | 0 | 10,312 | 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 | 128K | 1 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 128K | 2 | Valid | 138.5 | 69.3 | 600 ms | 14.4 ms | 2.0 | 0 | 4,154 | 30.0 s |
| Sustained decode | 128K | 4 | Valid | 228.6 | 57.0 | 834 ms | 17.5 ms | 4.0 | 0 | 6,856 | 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 | Valid | 68.1 | 68.4 | 915 ms | 14.6 ms | 1.0 | 0 | 2,043 | 30.0 s |
| Sustained decode | 248K | 2 | Valid | 88.2 | 44.1 | 1,169 ms | 22.7 ms | 2.0 | 0 | 2,644 | 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 | — | — | — | — | — | — | — | — |
Looped: the model kept repeating the same text, so the benchmark discarded this measurement. Over capacity: not run because this many requests at this context would exceed the server's KV cache.