Hello, I recently purchased an RX-580 4GB (Sapphire Nitro+). Previously I had a GTX 1050Ti, no other hardware has changed apart from a RAM upgrade. When replacing my graphics card, I made sure to DDU it first, but since this problem has been persisting I've done a full reinstallation of Windows and yet it still occurs.
The problem is, during games one of two things will happen, typically after 20-60 minutes - either the game will just crash to desktop (and the bug reporter will come up if it has one, interestingly), or in the worst cases - but significantly rarer - the entire system will hard lock, the monitor will lose signal and a power button reset will have to occur.
I have tried a lot, now my CPU is overclocked to 4GHz (Ryzen 3 1200) but it's been like that since the day I got it back in August 2017 and shows no errors in synthetics (Prime95, IntelBurnTest) or games. I have extensively tested my RAM with benchmark applications, the windows memory checking tool, and passmarks memtest tool, and there are no errors there. I even disabled XMP (3000MHz 15-15-15-35) and ran it at the standard JEDEC rate of 2133MHz, and these crashes still occured, plus I've had no errors in any intensive programs that would be typical of RAM.
For the card itself, I've tried a ton of different drivers and none of it makes a difference, I'm currently on 19.1.2. (I just bought the Resident Evil 2 remake), I have the power limit set to +50%, I've tried undervolting, and overvolting. The games I typically play are Civ VI, Insurgency Sandstorm, Project Cars 2, TES Skyrim Special Edition, MGSV:TPP.
My thoughts at the moment are that it's a VRAM issue, as very occasionally when booting, a grid of multicoloured (like noise) squares will appear on the screen in a checkerboard formation followed by a system hard lock, however, this only happens a) exclusively at startup and b) only after logging in to windows, I can wait for a million years on the login screen and do stuff, but nothing happens. I've done a ton of different VRAM tests, both dedicated utilities (memtestCL, Video Memory stress Test v1.7, MSI Kombustor artifact checker, OCCT artifact checker), and benchmarks (FurMark (with 8X MSAA), kombustor, Heaven, Superposition) yet not once has a crash ever happened, and I've let some of these run for an hour or more. It seems to only be when playing a game.
I also thought it could be my PSU, it's a 450W SuperFlower Golden Green HX, but I doubt it is this, as even with a +50% power limit the RX-580 can only use under the worst possible scenario (i.e. furmark) 260W, and my CPU under its worst condition (a R3-1200 at 4GHz @ 1.45v) can only use 100W. Neither component is ever near their limit, especially at the same time, and I did stress them both at their worst cases and it didn't crash, and this PSU gave no problems to the build with the 1050Ti.
In terms of temperatures, there is nothing wrong. My CPU under the heaviest loads (p95 and IntelBurnTest) maxes out at around 60-70C, and stays in the low 50s during games, my motherboard VRM stays in the 50s, and the card under a furmark load tops out at about 75C, lower in games.
Apologies for the lengthy post, but I'm genuienly out of ideas. I've been through the stickies, done the DCIM and SFC/SCANNOW, and had nothing to fix.