Skip to content

2x4ever/nvidia-fan-controller

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Nvidia GPU Fan Controller for linux

A simple standalone python script to keep your Nvidia GPUs below a given temperature, e.g.

python3 nvidia_fan_controller.py --target-temperature 60

This script uses a simple PID-controller to regulate the fan speed.

Dependencies

The nvidia_fan_controller.py script does not depend on any other python libraries. It does, however, require the following two command-line utilities:

  • nvidia-settings
  • nvidia-smi

Please make sure that these are installed on your system.

Also, make sure that you've enabled manual fan control on your system. This can be done by using the nvidia-xconfig command-line utility:

nvidia-xconfig --cool-bits=4

To update your /etc/X11/xorg.conf manually instead, scroll down to see an example.

Stress test

It is important to stress test your system to ensure that it won't catch fire 🔥

I found this simple stress test tool called gpu-burn. Here are the steps I use to run the stress test.

First let's clone this repo and the stress-test repo:

git clone https://github.com/KristianHolsheimer/nvidia-fan-controller.git
git clone https://github.com/wilicc/gpu-burn.git

Navigate to the nvidia-fan-controller dir and run:

python3 nvidia_fan_controller.py --target-temperature 60

In another terminal, navigate to the gpu-burn dir and compile (you only need to do this once):

make

then run (10 minutes = 600 secs):

./gpu-burn 600

You may want to open a third terminal to monitor your GPUs:

nvidia-smi --loop=1

If you'd like to selectively stress-test your GPUs, you can set the CUDA_VISIBLE_DEVICES variable. For example, here's how I sandwich an idle GPU between two fully-loaded ones:

CUDA_VISIBLE_DEVICES=0,2 ./gpu-burn 600

Install as a service

You would probably like to automatically start the controller at start-up. Here's how to do that using systemd.

It is highly recommended to a stress-test (like the example above) to check if the fan speed is properly regulated by the controller.

Create a systemd service file:

python3 nvidia_fan_controller.py --create-service-file

This creates a file called nvidia-fan-controller.service in the same dir as the main script. Check the contents of the service file to see if all fields are populated properly.

If all looks good, install the service using the following sequence of commands:

sudo cp nvidia-fan-controller.service /etc/systemd/system/
sudo systemctl enable nvidia-fan-controller.service
sudo systemctl daemon-reload
sudo systemctl restart nvidia-fan-controller.service

Uninstall service

If you wish to uninstall the service, run the following commands to undo the installation steps outlined above:

sudo systemctl stop nvidia-fan-controller.service
sudo systemctl disable nvidia-fan-controller.service
sudo systemctl daemon-reload
sudo rm /etc/systemd/system/nvidia-fan-controller.service

Example xorg.conf

Here's the xorg.conf that I use for my system. I've got three GPUs: one GTX 1080Ti and two GTX 1070Ti. This /etc/X11/xorg.conf file sets the Coolbits option to 4, which tells the nvidia driver to allow all fan settings to be set manually.

Section "ServerLayout"
    Identifier     "Layout0"
    Screen      0  "Screen0" Absolute 0 0
    Screen      1  "Screen1" Absolute 0 0
    Screen      2  "Screen2" Absolute 0 0
EndSection

Section "Monitor"
    Identifier     "Monitor0"
    VendorName     "Unknown"
    ModelName      "Unknown"
    Option         "DPMS"
EndSection

Section "Monitor"
    Identifier     "Monitor1"
    VendorName     "Unknown"
    ModelName      "Unknown"
    Option         "DPMS"
EndSection

Section "Monitor"
    Identifier     "Monitor2"
    VendorName     "Unknown"
    ModelName      "Unknown"
    Option         "DPMS"
EndSection

Section "Device"
    Identifier     "Device0"
    Driver         "nvidia"
    VendorName     "NVIDIA Corporation"
    BoardName      "GeForce GTX 1080 Ti"
    BusID          "PCI:1:0:0"
EndSection

Section "Device"
    Identifier     "Device1"
    Driver         "nvidia"
    VendorName     "NVIDIA Corporation"
    BoardName      "GeForce GTX 1070 Ti"
    BusID          "PCI:2:0:0"
EndSection

Section "Device"
    Identifier     "Device2"
    Driver         "nvidia"
    VendorName     "NVIDIA Corporation"
    BoardName      "GeForce GTX 1070 Ti"
    BusID          "PCI:3:0:0"
EndSection

Section "Screen"
    Identifier     "Screen0"
    Device         "Device0"
    Monitor        "Monitor0"
    DefaultDepth    24
    Option         "Coolbits" "4"
    SubSection     "Display"
        Depth       24
    EndSubSection
EndSection

Section "Screen"
    Identifier     "Screen1"
    Device         "Device1"
    Monitor        "Monitor1"
    DefaultDepth    24
    Option         "Coolbits" "4"
    SubSection     "Display"
        Depth       24
    EndSubSection
EndSection

Section "Screen"
    Identifier     "Screen2"
    Device         "Device2"
    Monitor        "Monitor2"
    DefaultDepth    24
    Option         "Coolbits" "4"
    SubSection     "Display"
        Depth       24
    EndSubSection
EndSection

Credits

The icon used in the github link preview is Fan by Amethyst Studio from the Noun Project.

Contributing

To check the script for inconsistencies, I run:

flake8 nvidia_fan_controller.py --ignore E501
mypy nvidia_fan_controller.py

About

Nvidia GPU Fan Controller for linux

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%