WindowServer intermittent crash (DumpGPURestart)

[Nermal]

I've seen a few scattered posts about this problem but haven't seen a dedicated thread so I figured it's time to make one.

When certain combinations of apps are running, WindowServer will occasionally crash and take the entire system down. In my case it's World of Warcraft plus Discord. I can run WoW by itself all day without issue, but once I have Discord open it can crash. Sometimes it'll go weeks without crashing, and sometimes it'll crash multiple times a night.

Once WindowServer crashes, everything on screen will freeze up, but the system itself is still responsive (I can still hear people on Discord and they can still hear me). But if I try to recover from the issue then a few things can happen:

- Occasionally it's possible to Cmd-Tab out of WoW, force quit it, then start it again. In this case it's likely to crash again within the next hour or so.
- Sometimes when attempting to Cmd-Tab out, the screen will go completely black. This doesn't seem to be recoverable.
- Connecting to the computer via ssh and asking it to restart seems to have about a 50% success rate.
- Most of the time I need to hold the power button until the computer turns off.

It's certainly not just me. I've seen other posts about the issue here on MR when using 3D apps (no WoW/Discord in the mix) but just in passing ("so it's not just me"-type posts).

The only workaround I've seen was a suggestion to turn of hardware acceleration in Discord, but this didn't help. I'm posting this from work so don't have log files handy, but can post some later (although the WoW thread linked above has some in it).

Any suggestions on what to do to fix this?

 

[wellander1]

What os and hardware are you using.?

 

[Nermal]

It affects at least 10.15.6, 10.15.7 and 11.0.1. Other posts indicate that the issue started in 10.15.2 but I can't confirm that myself (my previous computer only had 10.13 and it was never a problem there).

I'm on a 2020 iMac with 5700 XT, but there's a range of hardware in the linked thread.

 

[jgleigh]

I've got a MacPro 5,1 with a 5700 XT. I've seen the crash in both Catalina and Big Sur. In Catalina, it would sometimes recover on its own. In Big Sur, it hangs until something forces everything to quit and logout. Once back into WoW, everything is fine. Seems to always happen when I first login to my SL sanctum and pan the camera around.

 

[dontpokebearz]

Has anyone found any solution to this? I'm experiencing the issue, and interestingly enough across operating systems - macOS Mojave and Windows.

I'm wondering if a fresh install of Mojave / OpenCore would help or if this is a hardware problem.

(2009 Mac Pro, Mojave, Sapphire Pulse RX580 8GB)

 

[Nermal]

I haven't actually had the issue since OS 11.4. There was something in the release notes about resolving crashes with a different game but it may have been the same root issue.

If it's happening under Windows for you then it's possibly a hardware issue. I don't have Windows and have therefore never tested it, but it did appear to be a software issue as it apparently affected Arm systems too.

 

[dontpokebearz]

Hi Everyone, I want to update and say that I've found my issue - my thermal paste on my RX580 was baked. I scraped it off and reapplied thermal paste and it's working is it should. Working in both Windows and macOS.

I did some testing on a friends PC to see if my issue was isolated to just my Mac Pro, or the GPU. When launching Sea of Thieves on his PC it also crashed immediately. We tried under-volting and overclocking as some forums have suggested, but that didn't work.

Finally we checked thermal paste and reapplied. I was in shock that it worked, and gamed on it for 30 minutes, as well as stress testing.

If you bought your card used, check the paste!! Even if new, check the paste! (Note, if you remove the heat sink from the GPU, you'll need to reapply as well.)

Hope this helps!

 

[melakoi]

I'm thinking the same thing about baked thermal paste on my GPU as mine also intermittently froze and sometimes crashed, and other times logged me out. I'm using a 5,1 Mac Pro with 32GB ram, and NVIDIA GeForce GTX 680 4 GB. Not the greatest, but more than enough for music production. MY NVME SSD took a poop earlier this year and I found a 2.5" spinning laptop drive with mac OS on it, so that's what it's still using. I know it can't be that because even at 5,400 RPM, the original Mac Pros used a 1TB spinning disc, and they ran fast. Then I got a weird "dumpgpurestart" in activity monitor taking 93% of my CPU resources. I was like, that's odd, what's going on? That's what brought me here.

 

[Nermal]

I only just noticed your response, but in my case it was definitely a software issue. I haven't had it happen for years at this point, and I'm on the same computer.