Ugh! For F's sake. So, I've mostly been playing Oblivion Remastered, Indiana Jones before it, and did a complete run of RE Requiem, and things have been great. Something all of these games have in common is a range of modern RT effects, which can demand a lot of power from the GPU especially if also using DLAA.
As I mentioned in other threads, I've been undervolting my 5090s for some time, along with frame limiting to between 70-90 FPS @ 4K, and I often see max draw at ~250-350W, depending on the game, with some holding in the 380-460W range that have more RT effects. Well, recently, when testing CP2077, SW Outlaws, ME Exodus, and TW3 my 5800X3D rig with a Zotac TUF 5090 started doing the black screen/max fan crash. This was seemingly random other than the GPU hitting ~75c right before, I know this seems like a tell tale sign of something horrible on a hardware level.
Okay, so I can easily see where most folks will go with this. It's Zotac, it's 5090, it's NV drivers, connector, etc. Well, I checked hardware, cooling, connectors-yes, did an inspection on the 12VHPWR (looks beautiful for those in doubt), checked and rechecked hardware (gotta love that Lian Li mATX making it easy to take off all the panels for access), air cleaned the living sh out of it. Still had issues. The other random part of these maddening situations is that sometimes, during testing, everything would be fine, but when I stopped the game, got back to the desktop, then it would happen. I'd still be looking at MSI AB and see the GPU temps back in the 30-50 range, and bam, black screen with fans louder than when I forced them at 100%, during my testing, I tried keeping all 100%, thinking it might help, but again, nada. I felt like I was in that ST TNG episode where a probe is trying to reprogram the ship's computer and everyone is waiting for it to blow up like the USS Yamato.
Tried different drivers and even got to the point I thought MSI AB was the culprit and installed Zotac Firestorm, which, aside from lacking an OSD, wasn't as horrible as I thought it would be, but still didn't fix it. Hours of googling all the rabbit hole forums where most folks never found a solution, and all the theories folks had but couldn't solve, and then the dots finally started to connect in my head. What if none of the above is responsible, and it's the game launcher I'm using?
Well, launching games from Steam, NV App, and, choke, choke, Epic are working. Again, for F's sake, there was one that I never even considered was the evil gremlin behind it all; it turns out that all the games that were crashing were being launched via GOG Galaxy. Seriously, I couldn't believe it.
The nice thing about GOG is, of course, DRM-free, which also means I can run any of my games I've bought from there via their own EXEs directly. I tested my theory, and CP2077, TW3, and ME Exodus all ran flawlessly. Just for kicks, I also tested SW Outlaws and Hogwarts (since it has quite a bit of RT stuff as well), via the NV App, and not a single problem.
I know this experience isn't a cure-all solution for everyone, but it definitely is for me. Another plus from it is that while desperately trying to force various configurations to eliminate it, I found out I could wrench down the power limit all the way to 69% using AB, and those games which were hitting just above 450W are now hanging between 385-420 with the boost clock peaking just above 2400 MHz and mem at 29 GHz and still hit way over 120 FPS using MFG x3 and even lower temps (low-mid 60s, less fan noise.
