segfaults in chrome, libnvidia-glcore.so on GTX 1080 with 384.98 & 390.48

I have a Debian system with a Ryzen 1800X and GTX 1080. If I use google chrome with hardware acceleration enabled, I get regular segfaults in the browser perhaps every thirty minutes. Sometimes this manifests as a tab crash, and sometimes the entire system will freeze, continuing to display the same frame (sometimes with an audio loop) and not responding to network attempts to ssh in.

The strange part is that I have a very similar system with a Ryzen 1500X and GTX 1060 3GB, running 384.98, which never encounters crashes. At one point I copied the entire system image (dd) to the Ryzen 1800X, and immediately encountered crashes. I also get crashes with the latest 390.48 driver. It might be a coincidence, but perhaps there is some bug in support for 1080s?

lspci | grep NV

0e:00.0 VGA compatible controller: NVIDIA Corporation GP104 [GeForce GTX 1080] (rev a1)

Anyway, here are some of the types of segfaults I get.

grep segfault /var/log/messages

Apr 7 14:49:30 hostname kernel: [ 5470.561759] chrome[4772]: segfault at 607 ip 00005578c68f5b85 sp 00007fffd54d1fd0 error 6 in chrome[5578c221d000+5b98000]
Apr 7 15:09:18 hostname kernel: [ 881.517453] chrome[10135]: segfault at 1747ea4a205c ip 00007fd9c063bf01 sp 00007fff884e6520 error 6 in libnvidia-glcore.so.390.48[7fd9bf5d2000+141e000]
Apr 7 15:10:08 hostname kernel: [ 931.410787] TaskSchedulerFo[10585]: segfault at 7f9e1c2b2a56 ip 0000558bb8b3a083 sp 00007f9e1c2b3329 error 6 in chrome[558bb79ba000+5b98000]
Apr 7 15:49:49 hostname kernel: [ 3312.511249] Compositor[10665]: segfault at 558bc97569b7 ip 0000558bba2c3f07 sp 00007f9e29190b50 error 4 in chrome[558bb79ba000+5b98000]

nvidia-bug-report.log.gz (107 KB)

Did you check for the ryzen bug?
https://github.com/suaefar/ryzen-test

Ha yeah, I actually just RMA’d a Ryzen chip because it suffered from the segfault bug. This is a different problem because it manifests when the system isn’t under full load. But thanks for the suggestion, because it led me to the true problem: the system’s RAM couldn’t handle the speed it was overclocked at, causing system instability even on just CPU-bound applications. No GPU issues so far now that I’ve used more conservative settings for the RAM. Thanks again.