#2102 Community benchmark on NVIDIA A10

Descrição

# Community benchmark on NVIDIA A10

Measured on 2026-10-12 with `tools/strata_bench.py` (suite 1, 3 runs per test, medians with ranges). Speeds are the engine's own timings; first-token times are measured by the client over HTTP. Speed only: no correctness check was run.

## Hardware and software

- GPU and VRAM: NVIDIA A10, 22 GiB VRAM
- CPU: Hygon C86-3G 7380 32-core Processor (59 cores, 118 threads); installed RAM: 455.4 GiB; storage: not measured
- OS: Linux 4.19.90-52.58.v2207.ky10.x86_64; GPU driver: 590.48.01; CUDA: 13.1
- Strata: project 0.1.43, commit not a git checkout; engine 0.1.43
- Background workloads and power limits: not recorded

## Model and configuration

- Model and quantization: qwen3.8-flash-next-iq3_s
- Context: 262144 tokens; parallel slots: 1
- Settings: the install's own configuration, unchanged; thinking off for the suite; temperature 0

```text
strata --pack iq3_s --native Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S-00001-of-00002.gguf --ple-gguf Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S-00002-of-00002.gguf --expert-profile expert-profile.bin --expert-cache auto --prefill auto --spec 4 --spec-min-p 0.5 --mtp rt --max-context 262144 --kv int8 --ple-io ram --kv-resident 32768 --vision --vram-reserve-mib 700
```

## Method

Run by `python tools/strata_bench.py` against the HTTP API of a Strata server (already running); model loading is not timed. One warm-up chat and a 1K prompt are discarded first. Every prompt starts with a unique tag, so the conversation cache cannot help the fresh-prompt tests.

- short chat: the three questions of `tools/ab_engine.py`, up to 100 new tokens, 3 runs
- 4K / 16K prompts: generated word lists sized by a calibration request, 24 new tokens, 3 fresh prompts each
- agent session: system prompt and four tools, then 5 turns that each add about 1500 words of a project file and continue from the model's own earlier answers (the way agent clients resend a conversation), 64 new tokens per turn, 3 sessions; the cache numbers come from the answer's usage (`cached_tokens`) and timings (`prompt_n`)
- concurrency (when run): N requests sent at the same moment, total tokens over the wall time of the burst, latency over all requests

## Results

| Configuration | Actual prompt tokens | Reused tokens | Generated tokens | Runs | Prompt tok/s median and range | Decode tok/s median and range | TTFT seconds median and range |
| --- | ---: | ---: | ---: | ---: | --- | --- | --- |
| short chat | 46 (43-51) | 0 | 100 (100-100) | 9 | not measured | 17.1 (15.8-24.1) | 0.49 (0.41-0.84) |
| 4K prompt, fresh | 4019 (4019-4019) | 0 (0-0) | 24 cap | 3 | 1431.3 (1417.7-1433.8) | not measured | 2.84 (2.84-2.87) |
| 16K prompt, fresh | 16058 (16058-16058) | 0 (0-0) | 24 cap | 3 | 1709.0 (1707.9-1719.3) | not measured | 9.45 (9.39-9.46) |
| agent session, follow-up turns | 14465 (14446-14471) at the last turn | see below | up to 64 | 3 | see below | 19.4 (15.1-33.8) | 2.83 (2.26-208.34) |

Ranges are the lowest and highest of the runs; `Runs` counts requests that returned a token.

### Agent session (conversation cache)

Cache hit over the follow-up turns: 68.8 % (42.2-68.8); earlier-turn tokens read again per session: 4 (3-11602) (earlier turns that the cache did not hold; the 13634 (13633-13634) new tokens of the files are read once); first turn (nothing cached) first token: 1.26 s (1.26-1.26).

| Turn (run 1) | Prompt tokens | Reused | Read | Prompt tok/s | First token (s) | Decode tok/s |
| ---: | ---: | ---: | ---: | ---: | ---: | ---: |
| 0 | 609 | 0 | 609 | 496.4 | 1.257 | 11.5 |
| 1 | 3715 | 634 | 3081 | 1104.8 | 2.831 | 24.2 |
| 2 | 5248 | 3778 | 1470 | 648.8 | 2.31 | 19.1 |
| 3 | 8692 | 5311 | 3381 | 1178.8 | 2.917 | 18.2 |
| 4 | 11599 | 8739 | 2860 | 1029.9 | 2.832 | 33.8 |
| 5 | 14465 | 11619 | 2846 | 1025.7 | 2.833 | 17.5 |

Per-request data: `results.json`. Engine log lines from the start: `engine-log-excerpt.txt` (when this run started the server).

## Correctness and limitations

Not measured: this is a speed suite. The conversation uses the model's own replies, so token counts differ a little between runs. Machine details come from the server, `nvidia-smi`/`rocm-smi` and the OS; background load and power limits are not recorded. Total time of the suite: 393 s.