On 20/11/2025 16:00, Seb wrote:
Install latest firmware (BIOS). If it offers an option to run hardware self-test then try it.Only twice in 30 years have I upgraded the firmware for a motherboard. Followed instructions to the letter. No problem during the update. Both times it crippled the machine. I will only ever do that again on a machine that I am ready to discard.
At least read changelog between latest version and the version you have currently installed. A lot of things have changed over 30 years including expectations of vendors related to whether customers may promptly update firmware when some bugs are found.
I respect your choice, but it strongly affects my motivation to reply.
Check for errors, warnings, and all other messages related to graphics adapter output of the following command:sudo journalctl -b You may discover e.g. missed firmware files.Good idea. But... I don't know what a "missing firmware" line looks like here. And the output is both copious and very boring to read line-by-line.
Certainly reading logs require some experience. In the case of journalctl and color terminal, errors and warnings should be highlighted, so skimming through whole log of current boot and reading just errors should not take much time. By default journalctl uses less(1). This pager allows to search (or even to filter) by regex patterns. So "firmware" and "amdgpu" may be first keywords to try.
If X is unreliable then you may use virtual console or ssh login from another machine (install ssh-server if have not it yet).Yeah, but it's supposed to become a user-facing PC, not a server in a closet...
If your X session crashes frequently then reading logs from another machine may be more comfortable. X may freeze with no reaction to keyboard and mouse. This case remote debugging may be convenient as well.
Kernel crash stacktrace from logs is helpful to determine if your exact problem has been reported to the upstream bug tracker and whether it has been fixed at least in the development branch.