tools · sub-star
thermal_monitor
background daemon logging thermal curves to a local database.
readme
A terminal daemon that watches the machine: CPU load, package temperature and memory, sampled twice a second and written to a local database. Built to keep a water-cooled 5900XT and an air-cooled 9070XT honest, because a snapshot tells you nothing about thermals, a history does.
artifact
▬ gpu hotspot · ▬ cpu package
simulated thermal curves in the style of the logged data.
signal log
- one coherent sample per tick: the CPU is polled once and threaded through the derived readings, so a single logged row cannot disagree with itself.
- reads package temps from coretemp/k10temp where the platform exposes them, and models a curve from load where it does not, because Windows and WSL block the thermal zones. an estimate is labelled on screen and flagged in its stored row.
- logs every sample into a local database for long-term trend analysis rather than a live-only readout.
- CPU load is measured across the whole interval between ticks, instead of a blocking poll that saw one sixth of it.
- VRAM clock is modelled from system activity rather than measured; reading the real value needs PyNVML or PyAMDGPUInfo.
built with
PythonSQLite