#2430 Community benchmark: IQ3_S 32k on NVIDIA GeForce RTX 3080 Ti 12GB, Ryzen 9 5950X, 128GB DDR4

Descripción

# Community benchmark on NVIDIA GeForce RTX 3080 Ti

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 3080 Ti, 12 GiB VRAM
- CPU: AMD Ryzen 9 5950X 16-Core Processor (16 cores, 32 threads); installed RAM: 127.9 GiB; storage: not measured
- OS: Windows 10 (10.0.19045); GPU driver: 617.14; CUDA: not read
- 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: 32768 tokens; parallel slots: 1
- Settings: the install's own configuration, unchanged; thinking off for the suite; temperature 0

```text
strata.exe --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 --mtp rt --max-context 32768 --kv int8 --vision --vram-reserve-mib 700 --pcie-frac 0.00 --spec-min-p 0.70 --adapt-every 1 --adapt-swaps 160 --adapt-decay 0.97
```

## Method

Run by `python tools/strata_bench.py` against the HTTP API of a Strata server (started by the benchmark from the install configuration); 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 | 69.0 (64.9-81.0) | 0.52 (0.46-0.58) |
| 4K prompt, fresh | 4019 (4019-4019) | 0 (0-0) | 24 cap | 3 | 1864.0 (1659.2-1868.2) | not measured | 2.18 (2.17-2.45) |
| 16K prompt, fresh | 16058 (16058-16058) | 0 (0-0) | 24 cap | 3 | 2003.3 (1977.0-2003.9) | not measured | 8.05 (8.04-8.15) |
| agent session, follow-up turns | 14539 (14536-14574) at the last turn | see below | up to 64 | 3 | see below | 66.3 (54.5-87.1) | 2.14 (2.07-2.71) |

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: 68.6 % (68.6-68.8); earlier-turn tokens read again per session: 73 (2-74) (earlier turns that the cache did not hold; the 13729 (13728-13729) new tokens of the files are read once); first turn (nothing cached) first token: 1.34 s (1.32-1.38).

| Turn (run 1) | Prompt tokens | Reused | Read | Prompt tok/s | First token (s) | Decode tok/s |
| ---: | ---: | ---: | ---: | ---: | ---: | ---: |
| 0 | 609 | 0 | 609 | 449.0 | 1.378 | 53.4 |
| 1 | 3752 | 631 | 3121 | 1161.8 | 2.711 | 66.3 |
| 2 | 5285 | 3815 | 1470 | 640.4 | 2.32 | 63.0 |
| 3 | 8729 | 5278 | 3451 | 1331.1 | 2.622 | 60.4 |
| 4 | 11616 | 8754 | 2862 | 1133.0 | 2.556 | 69.0 |
| 5 | 14539 | 11640 | 2899 | 1224.1 | 2.402 | 55.6 |

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: 103 s.

[results.json](https://github.com/user-attachments/files/33437085/results.json)