#2367 Community benchmark: NVIDIA GeForce RTX 3080 Ti / 64GB DDR4

Beschreibung

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 7 5800X3D 8-Core Processor (8 cores, 16 threads); installed RAM: 63.9 GiB; storage: not measured
- OS: Windows 11 (10.0.26300); GPU driver: 610.88; CUDA: not read
- Strata: project 0.1.44, commit ed4773eb; engine 0.1.44
- 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.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 262144 --kv int8 --resident-experts --kv-resident 32768 --pcie-frac 0.55 --spec-min-p 0.70 --adapt-every 1 --adapt-swaps 80 --adapt-decay 0.97
```

Environment of the config: `STRATA_RESIDENT_HEADROOM_GIB=2`

## 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 | 50.4 (48.6-60.6) | 0.69 (0.58-0.80) |
| 4K prompt, fresh | 4019 (4019-4019) | 0 (0-0) | 24 cap | 3 | 984.6 (964.3-985.2) | not measured | 4.13 (4.11-4.20) |
| 16K prompt, fresh | 16058 (16058-16058) | 0 (0-0) | 24 cap | 3 | 1223.4 (1218.7-1226.1) | not measured | 13.16 (13.15-13.22) |
| agent session, follow-up turns | 14551 (14543-14571) at the last turn | see below | up to 64 | 3 | see below | 52.9 (37.7-63.8) | 3.90 (3.52-4.10) |

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 % (68.5-68.7); earlier-turn tokens read again per session: 74 (3-144) (earlier turns that the cache did not hold; the 13728 (13728-13729) new tokens of the files are read once); first turn (nothing cached) first token: 2.27 s (2.25-2.30).

| Turn (run 1) | Prompt tokens | Reused | Read | Prompt tok/s | First token (s) | Decode tok/s |
| ---: | ---: | ---: | ---: | ---: | ---: | ---: |
| 0 | 609 | 0 | 609 | 273.8 | 2.249 | 38.0 |
| 1 | 3750 | 630 | 3120 | 779.3 | 4.05 | 52.6 |
| 2 | 5283 | 3813 | 1470 | 410.2 | 3.612 | 46.3 |
| 3 | 8727 | 5276 | 3451 | 848.1 | 4.104 | 55.4 |
| 4 | 11653 | 8720 | 2933 | 745.4 | 3.976 | 63.8 |
| 5 | 14571 | 11672 | 2899 | 735.7 | 3.991 | 41.6 |

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: 163 s.