Issue with AMD GPU driver on Ubuntu 18.04
Issue with AMD GPU driver on Ubuntu 18.04
I moved to Manjaro but problems persist... YouTube frequently stops on Brave with the SIGSEGV error. I'm using kernel 5.9.3-1.
Use Chromium or Firefox for better performance. Brave isn't recommended.
When it crashes, relevant entries show up in `journalctl --catalog --boot`. If you're new to this, the interface works identically to Vim. To dive deeper, press Shift+G. Use arrow keys or PgUp/Down to navigate lines. For searching, type "/" and enter your query like "sigsev" or "brave". Move through results with N and Shift+N. If possible, check the logs after a full crash by running `journalctl --catalog --boot=-1`. Avoid AMDGPU-Pro drivers—they're only needed for OpenCL. Consider adopting markdown in forums.
I moved to Firefox now... The site runs smoothly except occasionally the OS does. I checked `journalctl --catalog --boot` but can't pinpoint the issue. I spotted some red warnings: kernel sp5100-tco disabled, kvm turned off by BIOS, SD 10:0:0:0 not found, and a strange pulseaudio error. The ALSA driver seems faulty, possibly a bug. Consider reporting it to the developers.
dmesg | grep error displays no output
dmesg | grep amdgpu shows server GLX version string: 1.4
dmesg | grep amdgpu reveals client GLX version string: 1.4
dmesg | grep GLX lists version: 1.4
dmesg | grep GLES shows profile versions: 1.4, 1.1, 3.2
OpenGL core profile version: 4.6
Compatibility profile version: 4.6
GLES1 profile version: 1.1
GLES version: 3.2
OpenGL core profile: Mesa 20.2.1
OpenGL ES version: 4.60 (Mesa 20.2.1)
Reattempt this now. It shouldn't trigger glxinfo. Everything else appears okay though. I'm also curious if you have a card that experiences similar problems like my MSI 5700XT. I own a MSI card that handles memory very closely. On Windows it runs smoothly, but on Linux it causes random freezes and restarts. Perhaps the Windows driver accounts for this since AMDGPU doesn't, no idea. For my card, changing the memory clock from 875 to 850 might help. Consider setting "amdgpu.ppfeaturemask=0xffffffff" in the kernel command line—either at grub during boot or in /boot/grub/grub.cfg permanently. You can add it alongside "quite splash." After that, try installing corectrl and see if crashes persist.
The BIOS/UEFI disabling KVM likely indicates the virtualization extensions are turned off. The sdd and lightdm alerts aren't concerning either. The color-safe notification doesn't matter here either. The watchdog alert appears unrelated to the problem. PulseAudio activity seems normal, as the bug reports don't match the situation. After your crash and restart, you ran `journalctl --catalog --boot=-1`? Typically, GPU issues leave a crash dump in journalctl, especially with Intel GPUs. I haven't experienced GPU crashes on my previous AMD machine, so I can't confirm. Edited November 17, 2020 by elsandosgrande “[…] reports I could find […]” → “[…] reports that I could find […]”.
The system detected changes in kernel settings and topology. It enabled GPU mode and adjusted device configurations. VRAM usage was reported at 8176M, with GART and TMZ features noted as disabled. Memory initialization succeeded, and audio components bound properly. A memory clock adjustment of 850 MHz was attempted following the crash.