I hope I can get some help here. I jumped to the AMD team recently and decided to buy an AMD Radeon RX 7600. My problem is that I tried to control the fan speeds, because the fans don’t start working until it reaches 60 degrees Celsius. I tried several software options, but none of them worked. I also tried modifying the kernel at boot like this:
I tried to modify those options with sudo, of course. When I tried to modify the fan speed with echo "200" > pwm1 , I received the error “bash: echo: write error: Invalid argument.” I know the values can range from 0 to 255 (stopped to full speed). When I tried to modify pwm1_enable from 2 to 1, it didn’t give me an error, but when I checked it again, the value was still 2. Additionally, manually enabling fan 1 with echo "1" > fan1_enable didn’t change anything.
I think it might have something to do with the GPU firmware, since AMD has the 0 RPM enable option by default. In CoreCtrl, I tried to enable the fan curve. It successfully modified the file “fan_curve” located in /sys/devices/pci0000:00/0000:00:01.0/0000:01:00.0/0000:02:00.0/0000:03:00.0/gpu_od , but the GPU’s behavior didn’t change.
I was thinking like last resort to just wired the gpu fans to the motherboard and try to control it from there, since I can’t write a python script and monitor the temperature I could try that… but it’s a new gpu and I don’t want to mess around with it.
Thanks, LACT didn’t work either. My theory… is related to the graphic’s card firmware. The only solution I see is connecting the fan cables to the motherboard… but I don’t know… that implies a lot of work, soldering and the risk of damaging the fans.
I’ve been on that journey with my RX 7800 and found LACT to be the only working fancontrol solution (I could find). However LACT only works above the 0RPM limit so it was not immediately apparent that it was working.
A little example what that means: If you apply 60% static fan speed that doesnt mean your GPU will ramp up to 60% now but if you do something so your card goes above the 0RPM temperature treshold the fan speed is now at static 60%.
What I did to get that working:
Install the latest fedora40 release (the testing one)
Open LACT and enable the service (as LACT will tell you) and overclock support
Reboot
Apply a static fan speed for testing
Reboot again (Not sure if necessary)
Launch up a game or benchmark to test
Set a fitting fan curve
Reboot again (Not sure if necessary)
Launch up a game or benchmark to test
This does not solve your problem that the fans only start up from 60 degrees but it is (at least for me) a working fan control software and I no longer have to suffer from AMDs standard fan curve.
Im on a somewhat fresh Fedora 40 with Gnome Desktop.
Thanks, I’m going to try this as soon as possible. My Fedora installation uses the KDE desktop environment. I can spare some space on a secondary disk to test this. If it works, I’ll replicate it on my main Fedora installation.
As you mentioned, the performance bottleneck will likely still be present.
I can’t believe that AMD’s standards for graphics cards seem so restrictive, even though the drivers are open-source. With my previous Nvidia card, I could easily adjust the clock speed (although I never needed to, as it ran perfectly fine and stayed cool).
Hi, I want to solve my own problem. It seems this issue will be fixed in kernel version 6.13. I wish I had the expertise to apply the patch myself, but I haven’t found a suitable tutorial. I gues I’ll have to wait until fedora 42 or if that version of the kernel is submited for fedora 41.
the fans don’t start working until it reaches 60 degrees Celsius
This is really quite normal, graphics cards have a higher operating temperature than CPU’s. When the temperature is too low, the card will run sub-optimally and will not reach the target performance. An average optimal GPU operating temperature is ~65-85°C, and the maximum operating temperature is around ~100-110°C.
Its not been released yet., current state is 6.13-rc3.
I would suggest that you wait for the announcement of the “kernel test week for 6.13”. Then install the test kernel and give feedback. If it works for you then just keep using it.
It’s mentioned in many sources online, but i don’t know if i can provide technical docs from AMD or otherwise at this point. In general, GPUs (like CPUs) are designed to perform optimally in a certain temperature range, within that range all the components with different electric and thermal characteristics will work together efficiently. While going colder is usually not a problem in the same way that overheating is, there is generally not much benefit in trying to force the card below it’s designed operating range.
Without intel, nvidia or amd docs i have to assume those sources are misinformed.
The frequency of operation is reduced when overheating, for example see the turbo mode of CPUs that use thermal head room to run faster until the chip gets too hot.
Clock frequencies are crystal controlled and do not change based on temperature. Software will change clock frequencies to adjust to load and temperature.
Right that’s a different argument, ‘too hot for too long’ is indeed problematic. which would pertain to sustained loads at high temperatures. For that it doesn’t really matter if the fan initially kicks in at 50°C or at 60°C, as was mentioned in the opening post.
Long story short, i would personally leave this up to the driver and not force the fan to kick in earlier than necessary, since there is likely no real benefit to be gained.