Ryzen crashing while idle

I did, yes. This is the new one.

Ok, I was able to get on the MSI page and check the BIOS version. I’m on 7B78v2E released 2020-06-22. Should I still try version 7B78v2D even though it’s older?

I think the point is to try latest stable for the generation of X470 + 3000 series cpu before new stuff for new generation was introduced. The good candidates would be around Agesa 1.0.0.4 (minimum allowed for your configuration) + 1-2 BIOS versions that maybe iron out some initial problems, that’s the gist, assuming BIOS is the problem.

You should be able to post links now.
I added you up to basic user TL1.
This might be easier with solving your problem.

If the new bios is not a beta bios.
Then it might be worth trying.

Thank you. I’m on the latest non-beta bios. Here is the event log: event.evtx - Google Drive

Like @agurenko said, I would stay with AGESA 1.0.0.4 (even it is older).
My problems began with AGESA 1.0.8.1, with a regular BIOS (not beta)

Yeah, 7B78v2D looks like a good choice. I wouldn’t go below 7B78v2C version though as there were several releases for memory compatibility improvements.

Fascinating, I’ve been tweaking stable system for last few days to see what was the ultimate solution that worked for my system and I finally narrowed it down today to… PSU Idle Current setting alone.

@Gandhi Can you confirm that you’re running at Typical Current for idle?

@thorn Well, I downgraded my bios from version 7B78v2E (AGESA 1.0.0.6) to version 7B78v2D (AGESA 1.0.0.4). However, it doesn’t seem to have made a difference. Crashes still happen frequently on Windows.

@agurenko I am running with Typical Current enabled.

I’ve noticed that recently the crashes on Linux have pretty much stopped (still get one once a week or so. Much, MUCH better than before) while on Windows they are constant depending on what I’m doing. Not totally sure why.

How is the SoC voltage after downgrade? Do you use Ryzen power plan?

Without any manual changes to it, it sits at 1.096V by default. As for Ryzen Power Plan, I didn’t set up anything by that name on Linux or Windows, so unless it’s enabled by default I don’t think so.

This is a power plan for windows that looks like need to be used, I might be wrong, but it’s close to schedutil cpu governor in linux or at least it’s a topology aware governor for windows. You get it by installing chipset driver from amd site and based on some read it performs much better. I’d try that since you’re saying linux is now more stable for you. It was decommissioned for 5000 series though.

https://bugzilla.kernel.org/show_bug.cgi?id=206903
Last comment (135)

Wow, did I have a bad morning, after running stable for weeks, I’ve got 3 restarts in a quick succession :frowning: Not to say I’m sad because it happened again, but even more - I never had more than 1 reboot in a power cycle. That’s without any changes to the configuration or any software updates in a last days.

I’ve added the amdgpu parameter to the kernel boot list as per BZ, lets see if that will make any difference. AMD is keep silently closing my tickets due to inactivity without coming back to me :frowning:

So I made sure all the chipset drivers were installed, and then rebooted the computer. After enabling the Ryzen Power Plan I tried using the computer a bit to see if I could get it to crash. Sadly, no luck. It crashed within minutes of making the changes. It may simply be time to reinstall windows.

I went to go add “amdgpu.ppfeaturemask=0xffffbffd” to grub just now and found “processor.max_cstate=5” listed under GRUB_CMDLINE_LINUX. I guess when I was troubleshooting a while back I put that in there to try and solve the issue and just forgot about it? Maybe that’s why Linux has been working better that Windows for a while now. Idk.

I have an interesting observation over the weekend after the worst week in terms of restarts so far: While week was full of restarts, the weekend went completely stable. I even did a full 4h+ memory test with memtest86 and a few hours of burnin test for cpu/memory/gpu and nothing.
This morning, after only 15-20 minutes of uptime - restart.

The only difference between work days and weekend are: VPN (openvpn) and Chrome browser. I don’t use Chrome for personal use and I don’t normally keep work VPN on, however sometimes I forget to disconnect at the end of work day, so it may stay on later in the evenings, hence pretty much everything points to the chrome usage.

Now, I don’t really know how to confirm that since logs never show any errors per se before the reboot, but I would assume @Gandhi you’re also using Chrome?

@agurenko No, I use Firefox. I’ve also started having crashes during gameplay. Granted, it’s just Vrchat, but it used to not happen during gameplay at all. Interestingly, gameplay crashes have on happened on windows so far (I play 95% of my games on linux now thanks to proton). Haven’t tested more demanding games on windows yet to see if it makes a difference.

Idk if it matters, but my vpn is also openvpn based (airvpn). Probably not since I have crashes whether it’s on or not. Also, I have c-states disabled in bios and grub. This definitely helped on windows, but linux hasn’t really changed. Gonna try waiting for the new bios version from msi and see if it makes a difference when it comes out. If it doesn’t… idk, I’ll go from there.

@Gandhi There is a new BIOS based on AGESA 1.2.0.0 released, for your board it’s a beta and stable for mine, I’m trying it since this morning, lets see if that makes a difference.