#3107 Community Benchmark: RTX 4050 Laptop (6GB VRAM) + 48GB RAM

Description

# Community benchmark on NVIDIA GeForce RTX 4050 Laptop GPU

Measured on 2026-10-19 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 4050 Laptop GPU, 6 GiB VRAM
- CPU: 13th Gen Intel(R) Core(TM) i7-13620H (10 cores, 16 threads); installed RAM: 47.7 GiB; storage: not measured
- OS: Windows 11 (10.0.26200); GPU driver: 610.74; CUDA: not read
- Strata: project 0.1.45, commit not a git checkout; engine 0.1.45
- Background workloads and power limits: not recorded

## Model and configuration

- Model and quantization: qwen3.8-flash-next-q2_0
- Context: 32000 tokens; parallel slots: 1
- Settings: the install's own configuration, unchanged; thinking off for the suite; temperature 0

```text
strata.exe --pack q2_0 --native Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00001-of-00002.gguf --ple-gguf Qwen3.8-Flash-Next-GSQ-RCO-Q2_0-00002-of-00002.gguf --expert-profile expert-profile.bin --expert-cache auto --prefill auto --spec 4 --spec-min-p 0.5 --mtp rt --max-context 32000 --vision --vram-reserve-mib 0 --kv q4_0
```

## 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) | 8 | not measured | 21.4 (17.4-30.0) | 1.55 (1.32-24.24) |
| 4K prompt, fresh | 3939 (3939-3939) | 0 (0-0) | 24 cap | 3 | 391.5 (380.7-393.0) | not measured | 10.15 (10.08-10.43) |
| 16K prompt, fresh | 15716 (15716-15716) | 0 (0-0) | 24 cap | 3 | 392.7 (391.1-393.6) | not measured | 40.13 (40.02-40.27) |
| agent session, follow-up turns | 14524 (14499-14560) at the last turn | see below | up to 64 | 3 | see below | 17.0 (12.6-30.3) | 14.75 (7.18-17.16) |

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 % (68.5-68.6); earlier-turn tokens read again per session: 73 (73-74) (earlier turns that the cache did not hold; the 13717 (13717-13720) new tokens of the files are read once); first turn (nothing cached) first token: 4.04 s (4.01-4.09).

| Turn (run 1) | Prompt tokens | Reused | Read | Prompt tok/s | First token (s) | Decode tok/s |
| ---: | ---: | ---: | ---: | ---: | ---: | ---: |
| 0 | 609 | 0 | 609 | 154.3 | 4.012 | 16.3 |
| 1 | 3741 | 621 | 3120 | 202.4 | 15.5 | 17.9 |
| 2 | 5274 | 3804 | 1470 | 207.2 | 7.18 | 15.2 |
| 3 | 8707 | 5267 | 3440 | 205.7 | 16.819 | 14.8 |
| 4 | 11584 | 8722 | 2862 | 200.9 | 14.331 | 21.2 |
| 5 | 14499 | 11600 | 2899 | 198.2 | 14.709 | 12.6 |

### Skipped and failed

- failed chat: {'type': 'server_error', 'message': 'the engine stopped unexpectedly (exit code 1); the next request restarts it'}

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: 531 s.