I made an interesting discovery, I'm not sure how many people are affected by it though...
I have the RX 6700 XT MECH 2X 12G OC card from MSI, and it's been crashing 6-12 times a day on me since February when I bought it.
The crash:
Monitor 1 goes black, almost as if it's not detecting any input from the PC all of a sudden. Monitor 2 stays active and so does Windows, but, only for about 15 seconds... After 15 seconds Windows completely freeze frames, and monitor 2 goes black after about 3-5 seconds once it's a still image. The system reboots, and both monitors are active again as if nothing happened.
I couldn't find a single possible cause for the sudden crash in Event Viewer. Everything logged has been tested and debunked. TPM, XMP, SAM, everything that could possibly be faulting wasn't the issue. The crashes persisted through countless changes and tests, even interrupted some of them, but still no logs generated of the actual cause.
That's when I noticed the only other persistent thing through all of this. “Default Radeon WattMan settings have been restored due to unexpected system failure”...
Which had me scratching my head, because I've never applied an overclock or under volt to the card. So I ended up in the Tuning tab of the Performance section to check out the "default" settings.
Without any user input on my end, default settings - with the exception of disabling auto update for drivers, Radeon was trying to boost my card to 2660 MHz. I thought this was a bit odd, considering the maximum stock boost clock was up to 2581 MHz according to AMD, and up to 2620 MHz according to MSI for this card in particular.
It was trying to boost my card beyond both listed maximums.
This prompted me to open the overclocking settings and adjust them, voiding my warranty with AMD and MSI to apply the proper stock settings.
I also added a more aggressive fan curve. I didn't like the default fan curve that had the fans hitting 80% when the card hit 90c. It shouldn't ever be hitting 90c.
It's been a few hours now, trying to recreate the crash... Stress testing the card in every way I can think of... Nothing. It doesn't crash at all, regardless of what I do on the system.
I can run Halo Infinite and OBS, side by side on my 2 monitors, record gameplay footage, and render a 1080p .webm with transparency layers in After Effects/Media Encoder while listening to music on Spotify.
So now I'm sitting here scratching my head wondering why AMD's own software was trying to boost my card beyond not only their listed maximums, but also the higher maximum that MSI has listed, without any input or permission from the user. How many cards have suffered from this? Is it a software problem? Or is it some kind of hybrid hardware/software problem where my card is telling the software it's capable of Y when really it maxes out at X and the software is only doing what it's told? How and why is this a thing?
And who do I talk to about my warranty? I shouldn't have to forfeit my ability to RMA the card for "overclocking" it to stock settings.
