#2104 Community benchmark : 7900 XTX, D4-2666, swift-1.5-iq2_xs, Strata 0.1.43

Description

# Community benchmark on AMD Radeon RX 7900 XTX

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: AMD Radeon RX 7900 XTX, 24 GiB VRAM
- CPU: 13th Gen Intel(R) Core(TM) i5-13500 (14 cores, 20 threads); installed RAM: 126.8 GiB(DDR4 2666); storage: not measured
- OS: Windows 10 (10.0.19045); GPU driver: not read; CUDA: not read
- Strata: project 0.1.43, commit 7695859a; engine 0.1.43
- Background workloads and power limits: not recorded

## Model and configuration

- Model and quantization: swift-1.5-iq2_xs
- Context: 131072 tokens; parallel slots: 1
- Settings: the install's own configuration, unchanged (hip backend); thinking off for the suite; temperature 0

```text
strata.exe --pack swift-iq2_xs --native Swift-Qwen3.8-Flash-Next-GSQ-RCO-IQ2_XS-00001-of-00002.gguf --ple-gguf Swift-Qwen3.8-Flash-Next-GSQ-RCO-IQ2_XS-00001-of-00002.gguf --expert-profile expert-profile.bin --expert-cache auto --prefill auto --spec 4 --mtp rt --max-context 131072 --kv int8 --kv-resident 32768 --pcie-frac 0.35 --spec-min-p 0.70 --pool-workers 5 --adapt-every 1 --adapt-swaps 80 --adapt-decay 0.97
```

Environment of the config: `STRATA_HIPBLASLT_TUNING=gfx1100-hipblaslt-100500.txt`

## 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 | 83.2 (52.1-108.5) | 0.35 (0.30-0.47) |
| 4K prompt, fresh | 4019 (4019-4019) | 0 (0-0) | 24 cap | 3 | 746.6 (385.5-747.8) | not measured | 5.42 (5.40-10.45) |
| 16K prompt, fresh | 16058 (16058-16058) | 0 (0-0) | 24 cap | 3 | 842.6 (759.7-853.1) | not measured | 19.09 (18.87-21.19) |
| agent session, follow-up turns | 14473 (14438-14485) at the last turn | see below | up to 64 | 3 | see below | 81.5 (65.4-131.0) | 5.10 (4.11-14.99) |

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 % (68.7-68.8); earlier-turn tokens read again per session: 4 (3-4) (earlier turns that the cache did not hold; the 13633 (13633-13634) new tokens of the files are read once); first turn (nothing cached) first token: 3.28 s (3.16-5.22).

| Turn (run 1) | Prompt tokens | Reused | Read | Prompt tok/s | First token (s) | Decode tok/s |
| ---: | ---: | ---: | ---: | ---: | ---: | ---: |
| 0 | 609 | 0 | 609 | 117.3 | 5.225 | 51.9 |
| 1 | 3703 | 623 | 3080 | 205.8 | 14.992 | 67.1 |
| 2 | 5236 | 3767 | 1469 | 170.1 | 8.659 | 67.7 |
| 3 | 8680 | 5299 | 3381 | 252.0 | 13.442 | 73.6 |
| 4 | 11595 | 8735 | 2860 | 243.8 | 11.763 | 75.7 |
| 5 | 14485 | 11639 | 2846 | 307.4 | 9.287 | 65.4 |

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: 237 s.