Random "Freezing" with AMD Ryzen 3900 and NVIDIA 3090

Ask Fedora
, , ,
1

My Fedora 43 KDE system with an R9 3900 and Nvidia 3090 constantly freezes. Sometimes I can hear the audio it was playing. Sometimes not. I even kept SSH connected but it becomes unresponsive at exact moment my system freeze.

I ran memtest overnight. No errors.

I used CPU and GPU stress testers. No problems.

My PC is chill, I monitor CPU and GPU temps on a side monitor constantly. Both always around 40-60.

In fact, I can leave my PC unused but turned on for days. When I start casually using it (web browsing, light coding, etc.), it freezes and requires a full restart. Sometimes I get only a restart with an MCE error about Bank 5.

My PC was working stable for more than a year until an update broke it, probably around kernel 6.8. I’ve also tried reinstalling Fedora to my system. Didn’t work. I’m currently on the latest, 6.19.9. Using latest Nvidia open/properity drivers.

I don’t think this is a hardware error. I dual booted with Windows 10 to see if blue screens would be more helpful, but Windows runs perfectly fine with no errors.

Besides, take a look at similar threads opened about this issue. There is even a gist repo. This seems to be a well-known issue where some users have had luck by disabling C-states and switching power profile to “typical.” But I’m not one of them. I also can’t change my hardware with current ram economy.

Is there a fix for this issue? Or should I suffer with Winslop?

https://gist.github.com/dlqqq/876d74d030f80dc899fc58a244b72df0

2

I don’t normally answer Nvidia questions… but as no-one else has…

What Nvidia drivers are you running and how did you install them?

3

Oh thanks, promise to swap Nvidia as soon as possible :>

I’m on 580.142 latest open drivers from Nvidia. I followed official rpmfusion Howto. Secure boot disabled.

So just
sudo dnf install akmod-nvidia
sudo dnf install xorg-x11-drv-nvidia-cuda

1 Like
4

I’ve never seen MCE errors that were not hardware problems. Most often RAM, but failing power supplies or electrolytic capacitors are also common. Some OS upgrades encounter previously undetected RAM issues due to changes in the way memory is used.
https://en.wikipedia.org/wiki/Machine-check_exception

Try running standalone memtest86+ overnight for several nights.

5

I have a Ryzen 5600x and a B450 MOBO.

My PC stuttered, froze and even crashed for no easy-to-explain reason, even on Windows 10, usually when using the browser and installing game updates.

Now I don’t have Windows installed on the NVMe anymore (a long story, with the problems brought by activating ReBar), but even before I did that, one thing helped my Fedora KDE install to not stutter anymore:

Create a 4gb SWAP partition on any SSD in your PC. For some reason there’s something that benefits from having it.

It’s a pretty cheap and easy thing to try, so might as well…

1 Like