The AMD graphics card drivers can change CPU settings in the UEFI [Update]

[Update 06.04.2022]

After the problem could be reproduced several times in various forums, AMD has now made a public statement on the part of Toms Hardware. Namely, that this behavior of the driver was not intended and a fix is ​​being worked on. Until then, in this particular scenario, the only way to do it manually is to use the Ryzen Master SDK Radeon Software Slimmer to remove and then import the driver. So again for your information. Those who do not run OC or UV, neither on the part of the CPU nor the graphics card, are not affected by this problem. But if you load a profile on the graphics card, you should check your UEFI afterwards. source here.

I stumbled across this fact when I had installed the Adrenalin 22.3.1, as always wanted to load my UV (Undervolt) profile for the 6800XT and the computer promptly asked for a restart. That was new, it wasn’t like that before. After the restart the first problems came.

Unreproducible blue screens, black screens, sometimes a massive freeze of the system.
To briefly outline my specs, Ryzen 5900x operated via UV, socketed on a potent X570 board, 3600 CL16 Ram, and the said 6800XT, also operated via UV. The rest shouldn’t matter.

Of course, at the beginning I suspected my graphics card or the new driver behind it. And I wasn’t that wrong, even if I wasn’t right. So I took the UV out of the graphics card completely and the bugs were still there. Only the graphics card and the CPU(?) got a lot warmer. As a next step I went back to the previous driver, didn’t even load my UV profile, same error. So it’s not the driver, is it?

What’s the plan now?

A bit of helplessness slowly spread and I had the suspicion that my graphics card could produce the error. So the 6800XT out and the old RX580 in. Same phenomenon.
So it’s not the graphics card. What now? The PCIe x16 slot? Try it and use a PCIe x8, same phenomenon. Is it maybe the ram? So I mean I only updated one driver. One stick out, same phenomenon, swapped with the other, same phenomenon. Other slots, you can guess.

What now? The NT? Relatively unlikely. So I search in the software (which I should have done in hindsight). Namely in the event log. And lo and behold, since the driver update, the event log spits out WHEA errors! yes how now? Is my CPU in Murs now or what’s going on here? So into the UEFI, something is fishy here.
For explanation. I have four profiles stored in the UEFI.

1. 1. One in which I only allow the CPU 65W,
   2. one where she can treat herself to 100W,
   3. one by adjusting the voltages of all cores via Curve Optimizer so that the clock rates run on stock, but with less voltage,
   4. one where the voltages have been adjusted via Curve Optimizer in such a way that significantly higher clock rates are possible and the processor is pushed to the stable limit, so to speak, as well as the limits of the motherboard being exhausted (I never use it, total play stuff, only what would be possible if ).

I just don’t have standard. The third profile is actually running 24/7 for me, and it’s stable.
But now I get into the UEFI and see that my settings have been completely overwritten. The curve is gone and the processor was pushed to its absolute limit, apparently in 50MHz steps, i.e. to an unstable limit. So I heard it like that, started Windows, loaded my UV profile, went back to the UEFI, loaded my 24/7 profile, and lo and behold, everything works!

AHA

And now I was curious. I took an SSD, installed Windows 11 on it, loaded the default settings in the UEFI beforehand, installed Adrenalin 22.3.1, loaded my UV profile…
…it was taken over without a restart. Even in the UEFI everything was the way it was afterwards.
Driver uninstalled, profile loaded in UEFI, driver installed, profile loaded, required reboot, WHEA error.

So tried the same game with the RX580. Here, for whatever reason, the UEFI is left completely alone. I then borrowed a 5700XT from a friend, the same phenomenon as on the RX580.
Conversely, armed with my SSD and 6800XT, I went to my buddy who has an X570 board and a Ryzen 3900x. Nothing, all totally unremarkable. Everything as it should be.

The problem ONLY affects the Ryzen 5xxx in connection with an RDNA2 card, and then only if changes were made manually in the UEFI regarding the CPU and if a GPU profile is to be loaded in the driver. After some research I found the same thing on various forums. The “problem” is also eliminated by itself by using via Radeon Software Slimmer Deselects the “RyzenMaster SDK”.

Despite everything, I find this “behaviour” corrosive, especially because it is not mentioned at all in the release notes. What the Radeon Software does there, without the knowledge of the user, is to trigger an “AutoOC” of the CPU, which is unstable in 99% of the cases. So it works…

Tested on all Agesa versions available to me (1.2.0.3 – 1.2.0.6)