Before RMTing this card I thought I'd give it one last shot here.
I built a brand new system (win10) with:
ASUS B550 mobo
Corsair 32gb RT ddr
Antec 650w PSU
So here is the issue: Every game I tried (on Steam) crashes within 5 minutes. I get either a driver timeout error and crash to desktop or a green screen crash with PC reboot. I tried every single driver both with software and driver only and all of them cleanly installed. No difference. I can reproduce the crash without fail instantly by changing certain settings in games. For example, switching vsync off in a gpu intensive game. Switching to a high resolution like 4k. Both those result in an instant crash. I can delay the crashing by lowering settings way down. In other words, the harder I try to push the GPU, the faster it crashes. (there is a caveat to that, which I will mention in a moment).
THINGS I KNOW: In the event viewer, every crash was preceded with the DWM restarting, so obviously something is causing that to fail, that leads to a crash. It only happens in games, not when idling or browsing. I could not reproduce the crash in only one game, which as it turns out is CPU bound. I did isolated stress tests of the CPU, GPU and memory and all them combined, works flawlessly. Moreover I slotted the card in a completely different system (also win10) and the exact same thing happens, so it is highly unlikely that it is a system error outside of the gpu and/or drivers. It is also not an overheating problem on any component.
THINGS I TRIED: Flashed BIOS, re-installed windows, tried a range of games. Also, I installed Linux on a completely different drive, tried three games on Steam, all of them crash, hence it is not a DX11 or Win10 issue per se. Slotted an older card into the new system, works flawlessly.
GUESSES: I should note that the GPU stress test completely pushes the GPU and hence the VOLTAGE <----, and this is where I have a clue. When monitoring in game with the overlay I notice the voltage and mhz spiking wildly. It would sometimes go as low as 50mhz and right back to 2000+ the next second, only to go back down to sub 100 the next second. In a different game where the GPU gets used much harder everything works fine, until I stop moving with a semi static image in front of me; the GPU then does the "throttling" and crashes soon after. It would appear that the "throttling" is extremely aggressive and that something cannot keep up. I should also note that I tried two different PSUs. No difference. I tried locking the GPU speed but the setting in the software appears to not do anything as I am unable in any way to stop the card from throttling down. The only other thing I can think of is that there is perhaps a flaw in one of the VRAM modules because as I said, I cannot reproduce the error in a CPU bound game that uses less than 2Gb. VRAM. The minimum I saw in every other game was just over 4Gb.
THINGS I KINDA TRIED: I messed about with a bunch of BIOS settings and tried everything I could find on the web, like: the TdrDelay setting in the registry, the power plan to performance, and on and on.
I have been a loyal AMD customer for 20 years now but this is driving me insane and I am seriously considering going team green. If anyone has any ideas I would much appreciate it.