2.5 KiB
Universal K80 Cooling
This is a universal cooling software, which uses an Arduino Uno, PWM fan, and Python 3. It was designed for use with a Tesla K80, but should work with any NVIDIA GPU.
This will determine the speed your fan (connected to your Arduino) should run at based on the highest temperature between your GPUs, then will send that speed to your Arduino.
Make sure to set the port in the .py file you use!
How to use
- Connect the Arduino to the fan according to this diagram:
- Set the port variable
- Instructions for finding the port are in the Python code.
- Upload
rx_speed.ino
to your Arduino Uno, then runtx_speed.py
.- Make sure not to run the Python file until the Arduino sketch is done uploading! It should only take a few seconds.
- Revel in your success!
- If you plan to use this constantly, make sure to create a service that starts the Python code on boot.
My setup
I use this 3D-printed fan adapter for my Tesla K80; I highly recommend it.
Also, I use this fan. I wouldn't recommend buying it, though. I just use it because it's the most powerful fan I already had. It's also pretty loud.
In the long-term, I'm planning to build my own front cover for my PC case (a Corsair 175R), and will have a spot for my Arduino and wiring in it.
Pictures will be added later once my setup is complete. Don't expect it to happen anytime soon, though.
Notes and Limitations:
- This currently only supports NVIDIA GPUs
- This currently only supports normal PWM, not delta PWM
- You may need to change some variables in the Arduino code depending on your fan's specifications.
- Try looking up a datasheet
- This was designed for an Arduino Uno, and should work on one. I don't know about running it on anything else, though.
- I'm busy, so a lot of the files are probably still broken. However, the main stuff is working.
Resources I used
- Python-Arduino communications library - pySerialTransfer
- Arduino PWM fan control - 25 kHz 4 Pin PWM Fan Control with Arduino Uno
- This is the basis of my arduino-fan diagram
- See NVIDIA GPU temperatures - GPUtil