I know there's like a million similar threads on here, but since AMD doesn't seem interested enough yet to address the issue, I'll just add this post to the pile of looming WattMan issues.
So I've got a fairly new system, just under a year old and I've had nothing but issues with my RX 6800 XT ever since day one. I'll be playing any game for a bit and then out of nowhere my computer would die, reboot with MOBO indicating GPU failure and be stuck in boot with a black screen, requiring a manual reset to work again. Each and every time when I boot it up afterwards, the same message pops up...
"Default radeon WattMan settings have been restored"
I've tried all manner of things to solve it, but to no avail:
The one and only temporary solution I've found is to use driver only installation. The second I tried a driver only install of Adrenalin the issue went away. I haven't had a single system crash since then. Before that WattMan could crash my system upwards of three time a day.
I'd like to use Adrenalin Software again since it has some nice and useful features in there, but I'm stay as far away as possible from it until WattMan is either completely removed, reworked (and QA tested), or there's an added option to manually disable it.
I think I am having a similar issues. Mine typically isn't black screen, and typically not during game play. Rather it is using a citrix client for work. I get BSODs saying things like "DRIVER IRQL NOT LESS OR EQUAL" OR "UNEXPECTED KERNEL MODE TRAP", and many others. When the PC reboots I have a Wattman error popping up. It is hard to believe that AMD would allow such a clearly faulty bit of software, but maybe there interested in losing market share?
I have Ryzen 5 Radeon RX580. I've been having these problems on and off for more than 2 yrs. It really blows. I'll give what you suggested a try.
I found something worth a try on YT, go into bios and change pci from auto to gen3/4 depending on your board.
My temporary solution eventually stopped working for me. I ended up opening a warranty ticket with my retailer and getting my card replaced. I haven't had this issue again since then.
The only other thing I did, which was not mentioned in my original post was to use two separate power cables from PSU to GPU. My system was initially built (retailer) using a single cable with pigtail. Switching to two separate cables did seem to improve my failure rate a bit, in conjunction with driver only install.
Hope you get it sorted out somehow...