A fan controller daemon for Linux to control your AMD graphics cards, built in Rust.
Maybe. Use at your own risk.
Yes.
For Ubuntu/Debian you can download a .deb binary from the releases page. Then install it with:
sudo dpkg -i amdgpu-fancontrol_0.1.2_amd64.deb
sudo systemctl enable amdgpu-fancontrol.service
Now skip to the configuration section.
On each release there is also a pre-built binary attached, which can be used, in theory, on other distributions. You will need to add your own systemd service or equivalent, and point it to a valid config file.
To build and install this under Ubuntu/Debian, you will first need Rust installed, then run:
cargo deb --install
sudo systemctl enable amdgpu-fancontrol.service
Configure your card if needed at: /etc/amdgpu-fancontrol/config.toml. You can find your cards (on Ubuntu/Debian, at least), at: /sys/class/drm/.
In theory multiple cards are supported, but I don't own multiple cards, so bon chance.
If you want to adjust the window used to decide whether the fan can adjust downwards, you can specify it in seconds. Default is 30.
cards = ["card0"]
measurement_window = 30
cards_path = "/sys/class/drm"
endpoint_path = "device/hwmon/hwmon0"
monitoring_path = "/sys/kernel/debug/dri/0/amdgpu_pm_info"The two paths get joined together with the card names, for example: /sys/class/drm/card0/device/hwmon/hwmon0.
Start the service:
sudo systemctl start amdgpu-fancontrol.service
This works well on my machine, which is running a Sapphire RX 580 Nitro under Ubuntu 20.04, but I am not sure what issues could occur with different setups. If something weird happens and you want to quickly restore hardware fan control, first disable the service so it doesn't start on next boot:
sudo systemctl disable amdgpu-fancontrol.service
Then you can either stop the running service with:
sudo systemctl stop amdgpu-fancontrol.service
which should restore hardware control (although it seems to result in the fans running higher by default), or alternatively you can manually restore it with:
sudo echo "2" > /sys/class/drm/card0/device/hwmon/hwmon0/pwm1_enable
Lastly, rebooting after disabling the service should restore everything to normal.
Tests currently have to be run serially due to file modification gubbins, this can be done either by setting:
RUST_TEST_THREADS=1
in your environment, and running cargo test, or by running the tests with:
cargo test -- --test-threads=1
Patches or questions are welcome. Please submit bugs or issues via the issue tracker.