#2420 Community benchmark: NVIDIA GeForce RTX 3090, i7-12700KF, DDR4 64GB, IQ3_S, 262k k8v4

Description

# Community benchmark on NVIDIA GeForce RTX 3090

Measured on 2026-10-14 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 GeForce RTX 3090, 24 GiB VRAM
- CPU: 12th Gen Intel(R) Core(TM) i7-12700KF (12 cores, 20 threads); installed RAM: 63.8 GiB; storage: not measured
- OS: Windows 11 (10.0.26300); GPU driver: 581.80; CUDA: 13.0
- Strata: project 0.1.44.1, commit not a git checkout; engine 0.1.44.1
- 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

## Method

Run by `python tools/strata_bench.py --long` 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 | 91.6 (60.8-102.9) | 0.37 (0.28-0.48) |
| 4K prompt, fresh | 4019 (4019-4019) | 0 (0-0) | 24 cap | 3 | 1148.6 (1146.5-1163.5) | not measured | 3.53 (3.47-3.54) |
| 16K prompt, fresh | 16058 (16058-16058) | 0 (0-0) | 24 cap | 3 | 2090.7 (2000.5-2106.9) | not measured | 7.73 (7.65-8.09) |
| 32K prompt, fresh | 32049 (32049-32049) | 0 (0-0) | 24 cap | 3 | 2141.0 (2096.3-2142.2) | not measured | 15.03 (15.01-55.15) |
| agent session, follow-up turns | 14544 (14526-14557) at the last turn | see below | up to 64 | 3 | see below | 101.7 (61.2-130.7) | 3.30 (3.03-3.42) |

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: 68.7 % (68.6-68.8); earlier-turn tokens read again per session: 5 (4-73) (earlier turns that the cache did not hold; the 13728 (13728-13730) new tokens of the files are read once); first turn (nothing cached) first token: 2.17 s (2.14-2.27).

| Turn (run 1) | Prompt tokens | Reused | Read | Prompt tok/s | First token (s) | Decode tok/s |
| ---: | ---: | ---: | ---: | ---: | ---: | ---: |
| 0 | 609 | 0 | 609 | 272.6 | 2.271 | 62.4 |
| 1 | 3743 | 623 | 3120 | 935.9 | 3.355 | 70.8 |
| 2 | 5276 | 3736 | 1540 | 509.8 | 3.05 | 99.6 |
| 3 | 8720 | 5340 | 3380 | 998.2 | 3.424 | 101.7 |
| 4 | 11626 | 8764 | 2862 | 864.4 | 3.357 | 100.0 |
| 5 | 14544 | 11645 | 2899 | 880.2 | 3.325 | 65.9 |

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: 205 s.