Skip to content

FRG-99: Read the GPU temperature - #90

Merged
Teczo merged 3 commits into
mainfrom
feat/FRG-99-gpu-temperature
Nov 17, 2026
Merged

Teczo merged 3 commits into
mainfrom
feat/FRG-99-gpu-temperature

Conversation

@Teczo

@Teczo Teczo commented Nov 17, 2026

Copy link
Copy Markdown
Owner

FRG-99: TeczoForge reads the GPU temperature, shows it, and remembers how hot each job made the GPU. No pausing yet (that is FRG-100).

  • backend/src/gpuTemp.ts: every 5 s, execFile runs nvidia-smi --query-gpu=temperature.gpu --format=csv,noheader,nounits with a 3 s timeout (path from NVIDIA_SMI_PATH, default nvidia-smi). A failure makes it "unknown" and is logged once. A value older than 15 s counts as unknown. It only reads: no GPU setting is changed.
  • GET /api/gpu answers { temperatureC, readAt, known }.
  • Each running job keeps the highest reading (maxGpuTempC), saved when it finishes. A job shorter than one reading gets the reading at its end. A board card that goes back to Ready forgets it, so its next run measures again.
  • Batch bar: "GPU 62 °C" or "GPU temperature unknown". Time test "Show times": a "GPU max" column per preset (recent finished jobs). Gallery job details: "Highest GPU temperature".
  • backend/.env.example lists NVIDIA_SMI_PATH.

Tested: /api/gpu = 42.6 °C while nvidia-smi said 42.64 °C. With a wrong path: one warning, no crash, "unknown". Typecheck passes in backend and frontend. No new dependency.

🤖 Generated with Claude Code

Teczo and others added 3 commits November 18, 2026 02:48
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…details

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@Teczo
Teczo merged commit 519701b into main Nov 17, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant