LACT icon indicating copy to clipboard operation
LACT copied to clipboard

Linux AMDGPU Controller

Linux AMDGPU Control Application

This application allows you to control your AMD GPU on a Linux system.

Screenshot Screenshot Screenshot

Current features:

  • Viewing information about the GPU
  • Power/thermals monitoring
  • Fan curve control
  • Basic overclocking

Currently missing:

  • Voltage control on Vega20+ GPUs
  • Precise clock/voltage curve manipulation (currently can only set the maximum values)
  • Multi-GPU system support Should work now

Installation

  • Arch Linux: Install the AUR Package (or the -git version)
  • Debian/Ubuntu/Pop_OS: Download a .deb from releases. Warning: it has not been tested heavily
  • Otherwise, build from source:

Building from source

  • Install dependencies:

    • Ubuntu/Debian: sudo apt install cargo rustc libvulkan-dev git libgtk-3-dev make
    • Fedora: sudo dnf install git gtk3-devel rust cargo vulkan-headers perl-core
  • git clone https://github.com/ilyazzz/LACT && cd LACT

  • ./deploy.sh

Usage

Enable and start the service (otherwise you won't be able to change any settings):

sudo systemctl enable --now lactd

You can now use the application.

CLI

There is also a cli available.

  • Getting basic information:

    lact-cli info

    Example output:

    GPU Model: Radeon RX 570 Pulse 4GB
    GPU Vendor: Advanced Micro Devices, Inc. [AMD/ATI]
    Driver in use: amdgpu
    VBIOS Version: 113-1E3871U-O4C
    VRAM Size: 4096
    Link Speed: 8.0 GT/s PCIe
    
  • Getting current GPU stats:

    lact-cli metrics

    Example output:

    VRAM Usage: 545/4096MiB
    Temperature: 46°C
    Fan Speed: 785/3200RPM
    GPU Clock: 783MHz
    GPU Voltage: 0.975V
    VRAM Clock: 1750MHz
    Power Usage: 38/155W
    
  • Showing the current fan curve:

    lact-cli curve status

    Example output:

    Fan curve:
    20C°: 0%
    40C°: 0%
    60C°: 50%
    80C°: 88%
    100C°: 100%
    

Reporting issues

When reporting issues, please include your system info and GPU model.

If there's a crash, run lact-gui from the command line to get logs, or use journalctl -u lactd to see if the daemon crashed.

If there's an issue with GPU model identification please report it here, include your GPU model and the output of cat /sys/class/drm/card*/device/uevent.

Alternatives

If LACT doesn't end up working for you, make sure to check out CoreCtrl.