Speed test matrix
gemma-4-12B-it-int8-AutoRound on H200 NVL 141GB
sglang · 0.5.6.post3.dev5791+gf7d4e44d8 · INT85 contexts (8K–123K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 123K, 128K
Hardware
- Type
- Discrete GPU
- GPU
- H200 NVL
- VRAM
- 141 GB
- OS
- linux
Engine
- Engine
- sglang
- Version
- 0.5.6.post3.dev5791+gf7d4e44d8
- Quantization
- INT8
- Backend
- cuda
- Repository
- github.com/sgl-project/sglang
- Build
- cuda-13.2
- Commit
f7d4e44d82ac
Run
- Served model
- gemma
- Revision
- main
- Benchmark tool
- llm-inference-bench v0.4.29
- Decode test
- Fixed duration
- Prefill mode
- Measured during decode warm-up
- Time per measurement
- 30 s
- Max output
- 4,096 tokens
Launch command
"$HOME/env_h200-intel-sglang/bin/sglang" serve --model-path Intel/gemma-4-12B-it-int8-AutoRound --served-model-name gemma --trust-remote-code --tp 1 --reasoning-parser gemma4 --tool-call-parser gemma4 --context-length 131072 --mem-fraction-static 0.861057 --enable-metrics --host 0.0.0.0 --port 8000 --api-key <redacted>
Prefill
prompt tok/s by context length8K context7,2551,131 ms TTFT
16K context6,3662,554 ms TTFT
32K context5,0086,463 ms TTFT
64K context3,52318,327 ms TTFT
123K context2,28254,306 ms TTFT
128K context2,21758,172 ms TTFT
Sustained decode
total tok/s · rows are context length, columns are simultaneous requests| Context | 1 | 2 | 4 | 8 | 16 | 32 | 64 | 128 |
|---|---|---|---|---|---|---|---|---|
| 8K | 100.492 ms TTFT | 200.2120 ms TTFT | 399.1172 ms TTFT | 777.3204 ms TTFT | 1,385.2399 ms TTFT | 2,146.1768 ms TTFT | Over capacity | Over capacity |
| 16K | 88.1129 ms TTFT | 178.9165 ms TTFT | 353.2233 ms TTFT | 692.3400 ms TTFT | 1,252.8756 ms TTFT | Over capacity | Over capacity | Over capacity |
| 32K | 70.0202 ms TTFT | 144.6273 ms TTFT | 290.3404 ms TTFT | 580.1777 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity |
| 64K | 49.7345 ms TTFT | 105.2491 ms TTFT | 214.3868 ms TTFT | Over capacity | Over capacity | Over capacity | Over capacity | Over capacity |
| 123K | 32.4726 ms TTFT | 69.41,206 ms TTFT | 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 | 7,255.0 | — | 1,131 ms | — | — | 1 | — | — |
| Prefill | 16K | 1 | Valid | 6,366.0 | — | 2,554 ms | — | — | 1 | — | — |
| Prefill | 32K | 1 | Valid | 5,008.0 | — | 6,463 ms | — | — | 1 | — | — |
| Prefill | 64K | 1 | Valid | 3,523.0 | — | 18,327 ms | — | — | 1 | — | — |
| Prefill | 123K | 1 | Valid | 2,282.0 | — | 54,306 ms | — | — | 1 | — | — |
| Prefill | 128K | 1 | Valid | 2,217.0 | — | 58,172 ms | — | — | 1 | — | — |
| Sustained decode | 8K | 1 | Valid | 100.4 | 100.7 | 92 ms | 9.9 ms | 1.0 | 0 | 3,011 | 30.0 s |
| Sustained decode | 8K | 2 | Valid | 200.2 | 100.6 | 120 ms | 9.9 ms | 2.0 | 0 | 6,004 | 30.0 s |
| Sustained decode | 8K | 4 | Valid | 399.1 | 100.4 | 172 ms | 10.0 ms | 4.0 | 0 | 11,972 | 30.0 s |
| Sustained decode | 8K | 8 | Valid | 777.3 | 97.9 | 204 ms | 10.2 ms | 8.0 | 0 | 23,320 | 30.0 s |
| Sustained decode | 8K | 16 | Valid | 1,385.2 | 87.1 | 399 ms | 11.5 ms | 16.0 | 0 | 41,552 | 30.0 s |
| Sustained decode | 8K | 32 | Valid | 2,146.1 | 67.5 | 768 ms | 14.8 ms | 32.0 | 0 | 64,385 | 30.0 s |
| Sustained decode | 8K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 8K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 1 | Valid | 88.1 | 88.3 | 129 ms | 11.3 ms | 1.0 | 0 | 2,642 | 30.0 s |
| Sustained decode | 16K | 2 | Valid | 178.9 | 89.7 | 165 ms | 11.1 ms | 2.0 | 0 | 5,366 | 30.0 s |
| Sustained decode | 16K | 4 | Valid | 353.2 | 88.7 | 233 ms | 11.3 ms | 4.0 | 0 | 10,596 | 30.0 s |
| Sustained decode | 16K | 8 | Valid | 692.3 | 87.0 | 400 ms | 11.5 ms | 8.0 | 0 | 20,766 | 30.0 s |
| Sustained decode | 16K | 16 | Valid | 1,252.8 | 78.7 | 756 ms | 12.7 ms | 16.0 | 0 | 37,584 | 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 | 70.0 | 70.1 | 202 ms | 14.3 ms | 1.0 | 0 | 2,100 | 30.0 s |
| Sustained decode | 32K | 2 | Valid | 144.6 | 72.4 | 273 ms | 13.8 ms | 2.0 | 0 | 4,338 | 30.0 s |
| Sustained decode | 32K | 4 | Valid | 290.3 | 72.8 | 404 ms | 13.7 ms | 4.0 | 0 | 8,708 | 30.0 s |
| Sustained decode | 32K | 8 | Valid | 580.1 | 72.9 | 777 ms | 13.7 ms | 8.0 | 0 | 17,400 | 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 | 49.7 | 49.7 | 345 ms | 20.1 ms | 1.0 | 0 | 1,490 | 30.0 s |
| Sustained decode | 64K | 2 | Valid | 105.2 | 52.6 | 491 ms | 19.0 ms | 2.0 | 0 | 3,154 | 30.0 s |
| Sustained decode | 64K | 4 | Valid | 214.3 | 53.7 | 868 ms | 18.6 ms | 4.0 | 0 | 6,428 | 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 | 123K | 1 | Valid | 32.4 | 32.4 | 726 ms | 30.8 ms | 1.0 | 0 | 972 | 30.0 s |
| Sustained decode | 123K | 2 | Valid | 69.4 | 34.8 | 1,206 ms | 28.7 ms | 2.0 | 0 | 2,082 | 30.0 s |
| Sustained decode | 123K | 4 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 123K | 8 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 123K | 16 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 123K | 32 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 123K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 123K | 128 | Over capacity | — | — | — | — | — | — | — | — |
Over capacity: not run because this many requests at this context would exceed the server's KV cache.