# Hardware capture, dev box (big-office), 3 Oct 2026

GPU: NVIDIA GeForce RTX 5090, 32607 MiB, driver 610.57.04
Power limit: 450.00 W (configured cap; not measured workload draw)
Clocks max SM: 3105 MHz; GPU temp at capture: 61 C
CPU: AMD Ryzen 9 9950X3D 16-Core, 32 logical, max 5756 MHz
RAM: 89 GiB total (70 GiB used at capture, Strata resident)
Kernel: 7.0.0-38-generic, Ubuntu
Python harness: 3.14.4
Engine: Strata v0.1.34 (binary self-report; checkout 1678de3, tag v0.1.34)
Model: Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S (79 GiB model dir incl. PLE split)
Serve args: --expert-cache auto, --prefill auto, --spec 4, --mtp rt,
  --max-context 131072, --kv int8, --kv-resident 32768, --pcie-frac 0.00,
  --spec-min-p 0.70, --vision, --vram-reserve-mib 700,
  --conversation-cache-mib 8192, --conversation-cache-slots 4
