Linux GPU Control Application

This application allows you to control your AMD, Nvidia or Intel GPU on a Linux system.

| GPU info | Overclocking | Fan control | | ----------------------------------| ----------------------------------| ----------------------------------| | | | | | Software info | Historical data | | | | | |

Features:

Detailed GPU information reporting
- Name and manufacturer
- VBIOS info
- VRAM info (Type/Manufacturer/Bus)
- Hardware unit info (CUs/SMs/EUs, ROP count)
- Resizable BAR status
- Vulkan features and extensions
Monitoring
- Configurable historical charts for power/thermals/frequency
- Throttling info
- Data CSV export
Power configuration
- Power cap
- Power states (AMD only)
Thermals configuration
- Custom fan curves (AMD/Nvidia)
- GPU firmware thermal options such as thermal and acoustic target/limit (AMD RDNA3+ only)
Overclocking
- GPU/VRAM clocks configuration
- GPU undervolting (via voltage offset on AMD, indirectly on Nvidia)
Settings profiles
- Automatic profile activation based on running processes or gamemode status
OpenTelemetry metrics exporter

GPU configuration is handled by a system service that does not depend on a graphical session (Wayland/X11).

The service can also be used standalone with a config file, for example in headless scenarios.

Quick links

Installation

Arch Linux: Install the package from official repositories: pacman -S lact (or lact-git from AUR for development builds).
Debian/Ubuntu/Derivatives: Download a .deb from releases.

It is only available on Debian 12+ and Ubuntu 22.04+ as older versions don't ship gtk4.
Fedora: use the Copr repository, or download an RPM from releases.
Bazzite/Fedora Atomic: Use the Flatpak.
Gentoo: Available in GURU.
OpenSUSE: an RPM is available in releases.

Only tumbleweed is supported as leap does not have the required dependencies in the repos.
NixOS: There is a package available in nixpkgs.
Flatpak (universal): Available on Flathub and in releases.

See the Flatpak documentation for additional notes.
Docker (service only, no GUI): See DOCKER.md
Build from source.

Note: Nvidia support requires the Nvidia proprietary driver with CUDA libraries installed.

Development builds

To get latest fixes or features that have not yet been released in a stable version, there are packages built from the latest commit that you can install from the test release or using the lact-git AUR package on Arch-based distros.

Note: the date that GitHub shows next to the test release is not when the packages were built, the actual date is specified next to the attached package files.

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 GUI to change settings and view information.

Hardware support

See the Wiki page

Configuration

There is a configuration file available in /etc/lact/config.yaml. Most of the settings are accessible through the GUI, but some of them may be useful to be edited manually (like admin_group and admin_user to specify who has access to the daemon)

See CONFIG.md for more information.

Socket permissions setup:

By default, LACT uses either ether the wheel or sudo group (whichever is available) for the ownership of the unix socket that the GUI needs to connect to.

On most desktop configurations (such as the default setup on Arch-based, most Debian-based or Fedora systems) this includes the default user, so you do not need to configure this.

However, some systems may have different user configuration. In particular, this has been reported to be a problem on OpenSUSE.

To fix socket permissions in such configurations, edit /etc/lact/config.yaml and under the daemon section either:

Set admin_user to your username
Set admin_group to a group that your user is a part of Then restart the service (sudo systemctl restart lactd).

Overclocking (AMD)

Some functionality requires enabling an option in the amdgpu driver, see the wiki page for more information.

Power profiles daemon note!

If you are using power-profiles-daemon (which is installed by default on many distributions), by default it may override the amdgpu performance level setting according to its own profile.

When using LACT 0.7.5+ and power-profiles-daemon 0.30+, LACT will try to connect to power-profiles-daemon and automatically disable the conflicting amdgpu action in ppd to avoid this conflict.

If running older versions, you can resolve this manually by creating a file at /etc/systemd/system/power-profiles-daemon.service.d/override.conf with the following contents:

[Service]
ExecStart=
ExecStart=/usr/libexec/power-profiles-daemon --block-action=amdgpu_dpm

Note: the /usr/libexec path might be different on your system, check it in systemctl status power-profiles-daemon

See https://github.com/ilya-zlobintsev/LACT/issues/370 for more information.

Suspend/Resume

As some of the GPU settings may get reset when suspending the system, LACT will reload them on system resume. This may not work on distributions which don't use systemd, as it relies on the org.freedesktop.login2 DBus interface.

Building from source

Dependencies:

rust 1.76+
gtk 4.6+
libadwaita 1.5+
git
pkg-config
clang
make
hwdata
libdrm

Optional Dependencies:

vulkan-tools
clinfo

Command to install all dependencies:

Fedora: sudo dnf install rust cargo make git clang gtk4-devel libadwaita-devel libdrm-devel vulkan-tools clinfo
Arch: sudo pacman -S --needed base-devel git clang make rust gtk4 libadwaita hwdata vulkan-tools clinfo

Steps:

git clone https://github.com/ilya-zlobintsev/LACT && cd LACT
make
sudo make install

It's possible to change which features LACT gets built with. To do so, replace the make command with the following variation:

Headless build with no GUI:

make build-release-headless

Remote management

It's possible to have the LACT daemon running on one machine, and then manage it remotely from another.

This is disabled by default, as the TCP connection does not have any authentication or encryption mechanism! Make sure to only use it in trusted networks and/or set up appropriate firewall rules.

To enable it, edit /etc/lact/config.yaml and add tcp_listen_address with your desired address and in the daemon section.

Example:

daemon:
  tcp_listen_address: 0.0.0.0:12853
  log_level: info
  admin_group: wheel
  disable_clocks_cleanup: false

After this restart the service (sudo systemctl restart lactd).

To connect to a remote instance with the GUI, run it with lact gui --tcp-address 192.168.1.10:12853.

CLI

There is also a cli available.

List system GPUs:

lact cli list-gpus

Example output:

10DE:2704-1462:5110-0000:09:00.0 (AD103 [GeForce RTX 4080])

Getting GPU information:

lact cli info

Example output:

$ lact cli info
GPU 10DE:2704-1462:5110-0000:09:00.0:
=====================================
GPU Model: NVIDIA GeForce RTX 4080 (0x10DE:0x2704)
Card Manufacturer: Micro-Star International Co., Ltd. [MSI] (0x1462)
Card Model: Unknown (0x5110)
Driver Used: nvidia 570.124.04
VBIOS Version: 95.03.1E.00.60
VRAM Size: 16376 MiB
GPU Family: Ada
Cuda Cores: 9728
SM Count: 76
ROP Count: 112 (14 * 8)
VRAM Type: GDDR6x
VRAM Manufacturer: Micron
L2 Cache: 65536 KiB
Resizeable bar: Enabled
CPU Accessible VRAM: 16384
Link Speed: 8 GT/s PCIe gen 3 x8

Profiles lact cli profile [COMMAND]
- List profiles:
  
  lact cli profile list
  
  Example output:
```
Default
Gaming
```

LACT

Install / Use

README

Linux GPU Control Application

Features:

Detailed GPU information reporting

Monitoring

Power configuration

Thermals configuration

Overclocking

Settings profiles

OpenTelemetry metrics exporter