I've opened this topic on Reddit but couldn't seem to get an exact problem identification. Maybe you guys could help me...
The GPU is AMD R9 380 and the motherboard is MSI H81M-P32 (MS-7846).
Not necessary to read this part:
So today I was playing Black Desert Online. I've decided to take some screenshots and turned on remastered mode, cranked up everything and disabled all optimizations. Few screenshots later, I went outside the room and forgot the game. I've heard some semi-loud noises coming but didn't realize it was my speakers. I came into the room a few minutes later just to see my pc went to a black screen and my speaker was blasting this stupid noise. This happened to me before with Black Desert but I usually just restarted the computer and switched the graphics to low so the GPU would cool off and nothing would happen.
Anyway, so my GPU was hot, I just restarted the computer to see white stripes and eventually BSOD.
I don't know GPU a lot but I didn't see any physical damage in the pins of PCI-e slot (both on GPU and on the motherboard) but here are the screenshots anyway:
GPU top, GPU bottom
I unplugged and plugged it two times (one for the photos) and on the last one, after an error and a reset, it started to work again. A few months ago, one of the ram slots broke down resulting in always BSOD. Also, I tried internal GPU when external wasn't working and it worked perfectly.
Update: So now I think my GPU or motherboard is somewhat damaged but I'm not sure how damaged it is. When I try to launch a game/furmark test, I get a black screen and I think the system crashes. When I hit the restart button on the case, the GPU fans go (i think) at full speed for no reason (at 40-45 degrees Celsius) and I don't get an output from the GPU until I use the PSU power switch...
I don't think it's a VRAM issue, I used a stress test on VRAM with VMST for a few minutes. I think it tests VRAM without rendering a 3D scene (also an odd thing, the test said at some point, it was using 1GB of VRAM but Windows 10 Task Manager said 0.4GB was being used). So the problem occurs when you try to render a 3D scene. Also no issue with FHD/4K video playback as well...
Update 2: Checked out a few webgl/canvas demos and benchmarks. No problems and quite good performance too.
Update 3: Tried a DirectX benchmark, after few seconds of benchmark, the system crashed again and now it has this yellow transparent tint at the place of those white stripes. I unplugged the GPU and those "burn marks" on the photos above are actually because of light. There aren't any kind of burn marks on the pins. Right now I unplugged PSU cables from the GPU and I'm using the internal GPU which is I think Intel HD Graphics 4600