Hey Shmerl, I'm indeed an early adopter and run one of the first generation Ryzen CPU's. Hence my rough experience with it; it took ages and uncountable BIOS updates till I had the computer running as intended. I never put the system on a real stress test but I assume that I'm just one of the unlucky ones who got his hands on one of the faulty CPU's. (My system doesn't freeze when on stress, but quite contrary, always when idling or doing easy tasks. Checking the internet for the problem it's obvious that I'm not the only one with a Ryzen CPU facing this problem.)

There are of course a few things that I could still try, like putting the system on a stress test. But sometimes I think about just going back to hardware that I used in the past and never faced such problems with.

The best option is to get a newer Ryzen. First generation Ryzens are just bad (I had exactly the same issue). But there is a workaround in various UEFIs, called "typical current idle", that disables some of the C6 states and it should help preventing the lockups.

See: https://bugzilla.kernel.org/show_bug.cgi?id=196683

Last edited by Shmerl on 5 Dec 2019 at 6:23 pm UTC

I've been putting my new RX 5700 XT through its paces for a few hours now. So far so good. Everything I've thrown at it thus far runs perfectly. The fact that I was already running Ubuntu's latest mainline kernel (5.4.2, simple and safe to install with Ukuu) and Mesa 19.3.0 rc6 from the kisak-mesa PPA probably made the switch less painful. I guess the experience wouldn't have been as good if I'd bought it even a few days earlier.

Backports of GPU hazard fixes to llvm9 is probably a major help, as for kernel, hopefully flip patch will land soon. According to Alex Deucher who submitted it, it should first land in Linus' master (future 5.5) and shortly after that to 5.4 patches queue.

Last edited by Shmerl on 5 Dec 2019 at 6:09 pm UTC

Quoting: ShmerlBackports of GPU hazard fixes to llvm9 is probably a major help, as for kernel, hopefully flip patch will land soon. According to Alex Deucher who submitted it, it should first land in Linus' master (future 5.5) and shortly after that to 5.4 patches queue.

I don't have most of the games mentioned in the report(s) you're referring to, but I can't seem to trigger the flip bug with the ones I do. Maybe I'm just lucky, but in that case I hope it lasts.

That's good, it didn't happen in every game, but at least a couple (Hedon and Pathfinder: Kingmaker) were affected by it for me. Plus, sometimes KWin itself triggered it.

The flip patch [landed in 5.5 tree by the way](https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/drivers/gpu/drm/amd/display/dc/dcn20/dcn20_hwseq.c?id=627f75d18910b287472593a4a2c41de9a386f5a2).

It will probably end up in 5.4.3 next weekend. As for powerplay, I'll be testing 5.5-rc1 once it comes out.

Quoting: PlintslîchoMy system doesn't freeze when on stress, but quite contrary, always when idling or doing easy tasks. Checking the internet for the problem it's obvious that I'm not the only one with a Ryzen CPU facing this problem.

My brother bought a Rizen 7 1800x and he had the idle crash, i installed [disable-c6-systemd](https://aur.archlinux.org/packages/disable-c6-systemd/) from the AUR, and no more idle crash, in his case it is probably because he use an old PSU that does not support C6/C7 state.

You have to be sure the service is started, otherwise it will do nothing, and also your CPU will always be at a higher clock, and you will lose the turbo.

That problem with C6 is not related to PSUs. It's just a hardware bug in the fist generation Ryzens. My PSU supports C6 just fine, but with Ryzen 1700X it was hanging, and that was solved only once I upgraded the processor.

Last edited by Shmerl on 8 Dec 2019 at 5:32 pm UTC

Just tried 5.5-rc1.

1. powerplay issues with "failed to send message" are gone.
2. Flicker when resuming from suspend bug is gone.
3. Flip patch is already there, so upstream has the fix for related GPU hangs in Pathfinder: Kingmaker and etc.

Very good progress there! So I'd say if you aim to buy a Navi GPU but don't want to deal with manual fixing, 5.5 should be already good out of the box, paired with Mesa 19.3 and llvm 10.

Last edited by Shmerl on 9 Dec 2019 at 1:21 am UTC

Quoting: sergeMy brother bought a Rizen 7 1800x and he had the idle crash, i installed [disable-c6-systemd](https://aur.archlinux.org/packages/disable-c6-systemd/) from the AUR, and no more idle crash, in his case it is probably because he use an old PSU that does not support C6/C7 state.

You have to be sure the service is started, otherwise it will do nothing, and also your CPU will always be at a higher clock, and you will lose the turbo.

Thanks guys, much appreciated.

I have for the time being reverted all boot loader as well as BIOS options since nothing has helped anyway. I may be looking into getting a new CPU next year. Let's see what next year will bring, if Intel will have something to compete with AMD's CPU's. And then there's the next generation of Ryzen CPU's and mainboards that are supposed to hit the market by the end of next year as well.

But for now I don't care much anymore. It's the end of the year, not much time for computing, let alone building a new one or replacing essential parts of it. For now I'll only look forward for the holidays. =)

Quoting: PlintslîchoI have for the time being reverted all boot loader as well as BIOS options since nothing has helped anyway.

So your UEFI didn't have the "typical current idle" option?

It has, but setting it or leaving it on auto doesn't make any difference.
So I set it back to auto now.

Same for my brother, bios update or changing bios options did nothing, neither kernel parameters.

But disabling C6 state with systemd is effective.

yeah i had the same bug with my early 1700X, bios option for Asus X370 prime didnt disable core c6 state but only package.

I just added this to startup:

zenstates --c6-disable

this will disable both:

C6 State - Package - Disabled
C6 State - Core - Disabled

Anyway i upgraded to 3700X and now i can have it enabled with no issues.

Yeah, first generation Ryzens are just too buggy. Save yourself the trouble and get a new Zen 2 chip (3000 series).

Quoting: ShmerlFor me it got fixed, but those messages still spam dmesg, so I disabled most amdgpu sensors meanwhile, except for temperature monitor in KDE plasmoid, which doesn't trigger it for me.

@Smerl, which sensors and where did you disable it? I have the same issues, so I hope your workaround will help me.

Fan and temperature sensors in ksysguard.

Running 5.5-rc5 with amdgpu sensors in ksysguard. Looks like the powerplay bug is fixed now.

Current Oibaf Mesa and Kernels 5.3.x (on 5.3.18 now) is still the only stable configuration with the RX 5700 for me.

With Kernel 5.5-rc5 I usually can't even reach the desktop, either it gets stuck during boot on black screen or error message "[drm:dm_helpers_parse_edid_caps [amdgpu]] *ERROR* Couldn't read SADs: -2". Or I reach the desktop, but then getting hard freezes.

So I guess I'm holding on to Kernels 5.3.x for now :|