Speed test matrix
gemma-4-E4B-it on RTX PRO 6000 Blackwell Server Edition 96GB
sglang · 0.5.22.dev236+g1093c501d · BF165 contexts (8K–123K) × 8 concurrency levels (1–128) · prefill at 8K, 16K, 32K, 64K, 123K, 128K
Hardware
- Type
- Discrete GPU
- GPU
- RTX PRO 6000 Blackwell Server Edition
- VRAM
- 96 GB
- OS
- linux
Engine
- Engine
- sglang
- Version
- 0.5.22.dev236+g1093c501d
- Quantization
- BF16
- Backend
- cuda
- Repository
- github.com/sgl-project/sglang
- Build
- cuda-13.2
- Commit
1093c501dfef
Run
- Served model
- gemma
- Revision
- main
- Benchmark tool
- llm-inference-bench v0.7.6
- 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_rtxpro6k-google-sglang-1093c501dfef/bin/sglang" serve --model-path google/gemma-4-E4B-it --served-model-name gemma --trust-remote-code --tp 1 --reasoning-parser gemma4 --tool-call-parser gemma4 --context-length 131072 --mem-fraction-static 0.78 --enable-metrics --host 0.0.0.0 --port 8000 --api-key <redacted> --attention-backend triton --bf16-gemm-backend torch
Prefill
prompt tok/s by context length8K context22.9K358 ms TTFT
16K context20.4K795 ms TTFT
32K context15.8K2,043 ms TTFT
64K context9,9876,460 ms TTFT
123K context5,91220,949 ms TTFT
128K context5,70722,584 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 | 97.3333 ms TTFT | Looped | Looped | Looped | Looped | Looped | Over capacity | Over capacity |
| 16K | 87.649 ms TTFT | Looped | Looped | Looped | Looped | Looped | Over capacity | Over capacity |
| 32K | Looped | Looped | Looped | Looped | Looped | Over capacity | Over capacity | Over capacity |
| 64K | Looped | 106.4195 ms TTFT | Looped | Looped | Over capacity | Over capacity | Over capacity | Over capacity |
| 123K | Looped | 74.0334 ms TTFT | 152.4429 ms TTFT | 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 | 22.9K | — | 358 ms | — | — | 1 | — | — |
| Prefill | 16K | 1 | Valid | 20.4K | — | 795 ms | — | — | 1 | — | — |
| Prefill | 32K | 1 | Valid | 15.8K | — | 2,043 ms | — | — | 1 | — | — |
| Prefill | 64K | 1 | Valid | 9,987.0 | — | 6,460 ms | — | — | 1 | — | — |
| Prefill | 123K | 1 | Valid | 5,912.0 | — | 20,949 ms | — | — | 1 | — | — |
| Prefill | 128K | 1 | Valid | 5,707.0 | — | 22,584 ms | — | — | 1 | — | — |
| Sustained decode | 8K | 1 | Valid | 97.3 | 97.6 | 333 ms | 10.2 ms | 1.0 | 0 | 2,917 | 30.0 s |
| Sustained decode | 8K | 2 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 8K | 4 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 8K | 8 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 8K | 16 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 8K | 32 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 8K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 8K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 1 | Valid | 87.6 | 87.8 | 49 ms | 11.4 ms | 1.0 | 0 | 2,628 | 30.0 s |
| Sustained decode | 16K | 2 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 16K | 4 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 16K | 8 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 16K | 16 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 16K | 32 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 16K | 64 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 16K | 128 | Over capacity | — | — | — | — | — | — | — | — |
| Sustained decode | 32K | 1 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 32K | 2 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 32K | 4 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 32K | 8 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 32K | 16 | Looped | — | — | — | — | — | 0 | 0 | — |
| 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 | Valid | 106.4 | 53.2 | 195 ms | 18.8 ms | 2.0 | 0 | 3,191 | 30.0 s |
| Sustained decode | 64K | 4 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 64K | 8 | Looped | — | — | — | — | — | 0 | 0 | — |
| 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 | Looped | — | — | — | — | — | 0 | 0 | — |
| Sustained decode | 123K | 2 | Valid | 74.0 | 36.9 | 334 ms | 27.1 ms | 2.0 | 0 | 2,218 | 30.0 s |
| Sustained decode | 123K | 4 | Valid | 152.4 | 38.1 | 429 ms | 26.3 ms | 4.0 | 0 | 4,568 | 30.0 s |
| 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 | — | — | — | — | — | — | — | — |
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.