PenguinBurner wurde in der Version v0.7.9 veröffentlicht.

PenguinBurner ist ein Tool zur Verwaltung und Optimierung von Grafikkarten unter Linux, mit Fokus auf Power Management, Undervolting und Performance-Tuning.

Im Zentrum des Updates stehen Verbesserungen beim Suspend/Resume-Verhalten, sicheres Auto-UV ohne Power-Telemetrie sowie Unterstützung für Multi-GPU-Systeme mit eigenen Profilen pro Karte.

• Add mobile RTD3 deep-sleep handling so the daemon avoids keeping a sleeping GPU awake (#40). Special thanks to @Christine1204 for the very long debugging session on mobile sleep-state behavior.
• Reapply and verify saved GPU state after suspend and resume (#41).
• Allow Auto-UV to continue safely when power telemetry is unavailable (#48).
• Support Flatpak and pip daemon installation and repair on Bazzite and other immutable distributions through the canonical /var/opt/penguin-burner/libexec/penguin-burnerd path (#51). Thanks to @LostFire93 for the debugging session.
• Add per-GPU profiles, filtering, Steam targeting, and serial startup application (#46). Active daemon monitoring, adaptive switching, drift recovery, and fan control still cover one GPU at a time.
• Attribute RTD3 device-node activity to the target GPU on multi-GPU systems, so a process using another card cannot start or keep alive the target’s deferred runtime profile.
• Lower every GPU-table Performance target proportionally (the RTX 5080’s 2980 → 2950 MHz ratio, about 1%) so the default four-bin rising tail stays below each family’s clock ceiling; before this it could top out above the ceiling on some GPUs.
• Default the Balanced power limit to full board power like Performance, so the full scan reuses the balanced descent for the Performance tier and only runs its Auto-OC climb. Cap any tier per run from the scan dialog; only Efficiency stays capped by default.
• Daemon service install/uninstall/migrate commands ask for authorization themselves (pkexec/sudo) when not run as root, fixing the ModuleNotFoundError a bare sudo hit on pip installs.

Quelle: GitHub