b00nz/mVolt v0.45

on GitHub

v0.45 improves resilience to NVIDIA management-library failures and brings the command line up to date.

Reliability

• Isolated NVIDIA management (NVML) reads and GPU clock range calls in helper processes, so an NVML crash does not close mVolt+.
• Moved periodic clock reads and reconnect probes off the interface thread to keep the window responsive while NVML is slow or unresponsive.
• Added longer initialization timeouts, increasing retry delays and a retry limit for repeated failures. Closing the app cancels pending helper waits.
• Preserved an unknown state when a clock-setting request loses its response. These requests are never automatically repeated; an explicit Reset or reboot resolves the uncertain state.

Command line

• Added --thermal-input-1, --thermal-input-3, --thermal-input-4 and --thermal-input-5. Each accepts a fixed temperature in Celsius or default to restore the GPU-provided input.
• Added thermal-input state and requested, enforced and default power limits to --status, using the isolated NVML backend for power readings.
• Added stable adapter IDs to --list-gpus and the selected adapter ID and index to --status.
• Updated command-line help, examples and the guide, including the distinction between Default and a fixed 0 °C input.