Speed test matrix
internlm_internlm3-8b-instruct-autoround-int8-gs128-sym on RTX 3090 24GB
vllm · 0.32.0 · AutoRound INT8 W8A16 gs1283 contexts (0–24K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 24K
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
- AutoRound INT8 W8A16 gs128
- 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), 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 fbaldassarri/internlm_internlm3-8b-instruct-autoround-int8-gs128-sym --max-model-len 32768 --gpu-memory-utilization 0.94 --max-num-seqs 128 --enable-prefix-caching --trust-remote-code --hf-overrides {"quantization_config":{"quant_method":"gptq","bits":8,"group_size":128,"sym":true,"desc_act":false,"lm_head":false}}Prefill
prompt tok/s by context length8K context4,5412,208 ms TTFT
16K context3,7655,278 ms TTFT
24K context3,2239,219 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 | 78.433 ms TTFT | 156.625 ms TTFT | 308.333 ms TTFT | 594.937 ms TTFT | 1,107.261 ms TTFT | Over capacity | Over capacity | Over capacity |
| 16K | 70.471 ms TTFT | 132.7111 ms TTFT | 246.0183 ms TTFT | 412.8279 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity |
| 24K | 66.695 ms TTFT | 124.3160 ms TTFT | 223.8240 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 | 4,541.0 | — | 2,208 ms | — | — | 1 | — | — |
| Prefill | 16K | 1 | Valid | 3,765.0 | — | 5,278 ms | — | — | 1 | — | — |
| Prefill | 24K | 1 | Valid | 3,223.0 | — | 9,219 ms | — | — | 1 | — | — |
| Sustained decode | 0 | 1 | Valid | 78.4 | 78.4 | 33 ms | 12.8 ms | 1.0 | 0 | 2,350 | 30.0 s |
| Sustained decode | 0 | 2 | Valid | 156.6 | 78.4 | 25 ms | 12.8 ms | 2.0 | 0 | 4,698 | 30.0 s |
| Sustained decode | 0 | 4 | Valid | 308.3 | 77.4 | 33 ms | 12.9 ms | 4.0 | 0 | 9,248 | 30.0 s |
| Sustained decode | 0 | 8 | Valid | 594.9 | 75.1 | 37 ms | 13.3 ms | 8.0 | 0 | 17,847 | 30.0 s |
| Sustained decode | 0 | 16 | Valid | 1,107.2 | 70.2 | 61 ms | 14.2 ms | 16.0 | 0 | 33,200 | 30.0 s |
| Sustained decode | 0 | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 0 | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 0 | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 1 | Valid | 70.4 | 70.4 | 71 ms | 14.2 ms | 1.0 | 0 | 2,110 | 30.0 s |
| Sustained decode | 16K | 2 | Valid | 132.7 | 66.4 | 111 ms | 15.1 ms | 2.0 | 0 | 3,980 | 30.0 s |
| Sustained decode | 16K | 4 | Valid | 246.0 | 61.6 | 183 ms | 16.2 ms | 4.0 | 0 | 7,380 | 30.0 s |
| Sustained decode | 16K | 8 | Valid | 412.8 | 51.5 | 279 ms | 19.4 ms | 8.0 | 0 | 12,376 | 30.0 s |
| 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 | 24K | 1 | Valid | 66.6 | 66.6 | 95 ms | 15.0 ms | 1.0 | 0 | 1,998 | 30.0 s |
| Sustained decode | 24K | 2 | Valid | 124.3 | 62.2 | 160 ms | 16.1 ms | 2.0 | 0 | 3,728 | 30.0 s |
| Sustained decode | 24K | 4 | Valid | 223.8 | 56.0 | 240 ms | 17.8 ms | 4.0 | 0 | 6,712 | 30.0 s |
| Sustained decode | 24K | 8 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 24K | 16 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 24K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 24K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 24K | 128 | Over capacity | — | — | — | — | — | — | — | — |
Over capacity: not run because this many requests at this context would exceed the server's KV cache.