Linux @lemmy.ml KokusnussRitter @discuss.tchncs.de 7h ago

[Solved] AMD GPU erratic fan behaviour

My GPU's fans spin either at 100% or stand still. they rarely ever dip below 1000rpm, and if so, only shortly (see images below). This worries me, and I am 99% sure I'll refund that card, but before buying a new one, I want to make sure the fault is not with my system. This is my first AMD GPU and I am pretty green in Linux.

So far I tried adjusting the GPU's fan curve, with fan-control and CoreCtrl but it didn't change the erratic fan behavior.

I checked my BIOS for 0rpm mode, but couldn't find anything. It also only let's me configure CPU and sys fan curves.

Is there anything I can try before sending the card back to fix these issues or -best case- verify the card is actually causing the issue? I'd hate to get a replacement and face the same issues again.

specs

Linux Mint GPU: acer AMD RX7600XT CPU: AMD Ryzen 5800x RAM: 16GB DDR4 @3600mHz; G.Skill Tridentz Neo MoBo: MSI B550Pro-A

images

Temperature and fan curves over the span of 20min; green = fan rpm, yellow is GPU junction temp, blue is GPU edge temp

Temperature and fan curves over the span of 20min; green curve is fan rpm, yellow curve is GPU junction temp, blue curve is GPU edge temp

Update I REMed it and will get the same.card again, but from a different, more widely adopted vendor. I don't want to spent another week troubleshooting an issue I may not have the skill to fix.

Thank you all for your input :)

19 comments

You may have just gotten a card with a bad fan voltage regulator (only works at full voltage). If you're positive they only work at full, then this is a potential symptom of either of those two things. The card itself is probably just fine.

Don't just test the automatic fan curve based on temp threshold, test to see if they work at a static 25/50/75% speed as well. If they really only work at 100, then definitely get an RMA if you're noticing heat issues.
- Yeah, set the fans to 50/75/100 and they still had 0rpm. Did not expect that to happen at 100% tbh. I'll try it again with other software, but I guess I'll RMA.
Which exact model of 7600 XT do you have?
- Yeah, he didn't share any information about the brand or vendor.
Off-topic but:

Linux Mint
GPU: AMD RX7600XT
CPU: AMD Ryzen 7 5800x
RAM: 16GB DDR4

That's surprisingly mid-range, I don't think I ever see that; just budget specs or bragging-level specs. As someone with almost the same build (Ryzen 7 4700GE, RX7700XTX, 16 GB DDR4) I'm positively surprised!
- It's the game development on a budget PC xD
  
  Needs no bling but to get stuff done.
  
  Honestly, I could do and have done Unreal, Unity and (easily) Godot development on less!
  
  But that's the minimum to spend for a very comfortable experience, I'd say; RX7900 and GTX 4090 are only there for bragging, while an on-board GPU would not be comfortable running game graphics in my opinion but I've seen people do it and not complain :/
Some vendors also have multiple BIOS-es on the same GPU card (like one overclocked BIOS, and the other is a "normal" BIOS).

You could check if there is a physical switch to switch between modes on your GPU. And see if the other GPU BIOS is also behaving like this.
Have you tried using LACT? https://github.com/ilya-zlobintsev/LACT

I have been annoyed for years but this is what saved my sanity when it comes to configuring GPUs on linux.

Also what specific model of card made by which company are you using? Fan curve config is something handled by each third party company differently.
- I took a bit of a trustfall and bought an acer rx7600xt. Plan to get the same chip, but from a different, more reputable vendor next.
  
  But I didn't try Lact, thanks. I'll give it a shot.
  
  It probably is in mints repositories but you will figure it out... But yeah that card shouldnt have any issues to stay cool in idle without much fan spin. Its normal for the fans to spin up pretty hard when gaming but if its during idle too then thats odd. You should add GPU usage to your psensor chart and add a screenshot of that.
1548 RPM should be slow for a small GPU fan, no? My Nvidia 3090 behaves exactly the same, switching the fan on and off as it hovers around 60C or so.

Looks like it’s working fine to me.

Also, take linux GPU monitors with a grain of salt. It’s possible the GPU fan RPM measurement is totally borked, and it basically represents "on" or "off." Check it with your eyes and ears instead, see if the fan is screaming or not. It shouldn’t be below 76C (as modern GPUs are configured to operate above 80C or so).
If the fan doesn't work correctly out of the box, it means that it needs driver or kernel support for the specific system. That's why for example, Tuxedo computers have extra drivers for the fans of their laptops. Back in the day, this wasn't an issue, as fans were kinda simple, and they were working on their own. These days, are software controlled, so it requires OS support.
- Do you have a recommendation where to look for fan drivers? In the package manager/ apt or do I have to hope the GPU manufacturer offers them?
  
  For the CPU fans there's not a good answer, but for the GPU, it should be with the driver's of the card. It should just have worked. Maybe you got a bug.
Imo it's hotspot temp. At least on my 2080ti if my hotspot reaches 104C (from thermal paste pump out) then the blower fan will go 100% which is insanely loud like a hair dryer.

A repaste will fix it till it pumps out again so next time I'm using a phase change material.
CoreCtrl fixed all my issues for the GPU and CPU fan.
I never had this issue with any of my AMD system I have a RX 6600 and a RX 7800XT systems both running different Linux distros

19 comments