The mCode
I can’t upload it directly as I’m a new user just for this reason. But I am using the latest version of the MCE from like two days ago. It at least identifies that the code exists, but it determines incorrect checksum and that it must be corrupt. This is the case for most Z790 BIOS I’ve found. I’m not for certain what’s going on, but it appears to have some data in the padding area that it’s not registering. That’s purely speculation though.
Edit: Sorry. I didn’t realize that I had the file locked. The link should work correctly now.
Could it be possible that it extends to other CPUIDs as well? If so, the CPUID would be excessively long which might account for the supposed corruption if it exceeds the expected value length. Though, I’m not programmer.
Interesting, I had already seen that corrupt microcode in the past but had discarded it as a mistake from Asus. The only change is at the headers, where “B0” was altered to “FF”:
It is interesting that they also altered the CPUID at the “extra” header, I was under the impression that this field is protected by the RSA signature as it is part of the RSA block:
Maybe only some fields of the “extra” header are protected and not all. This can be validated only by attempting to load that microcode and check whether the CPU microcode loader rejects it or not (which Asus probably did).
So, you’re saying that this is a “modded” older microcode which allowed a certain behavior that Intel blocked in subsequent releases.
In that case, I can mark it as a known “modded” microcode in MCE, and it won’t show the warning message, just a note that it is User/OEM modified.
Actually, from what I can tell, the microcode is valid Intel microcode that’s used on various different board manufacturers’ boards. I pulled BIOS files from two different manufacturers lately and the CRCs match. So they’re sourced from the same location. I pulled Gigabyte and ASUS.
Also, these were pulled from Z790 boards. So it’s not non-Z. And it also extends to 14th gen CPUs as indicated here: Intel CPUID Format v008
And this may be of use in figuring out why the microcode is listed as corrupt: Intel Microcode Caused “Unsupported CPU” BSOD Issue
No, these are not related. Intel would not push a microcode with an invalid CPUID (i.e. FF0671) and broken Checksum. That is usually, as dsanke said, a very clear indicator of OEM modification to accomplish a specific goal. This type of modification has been done in the past to allow non-K overclocking, AVX-512 support, BCLK alterations etc. For example:
306C3 → 99
506E3 → FF
90672 → FF
90675 → FF
The vendors either trick the OS to not load a newer microcode during boot (e.g. version FF) or implement BIOS-level loading in which a different microcode blob gets picked up according to a feature/setting, installed CPU etc. The latter is what is probably happening here. These types of modifications tend to propagate to other similar vendors so that they also offer the same capability. It is not weird for this to start with Asus and then find it itself at MSI, Gigabyte etc.
EDIT: As explained here, the 104 microcode is used to allow undervolting without performance loss. This confirms that it is a “modded” microcode for a specific capability no longer offered/allowed by Intel. Asus decided to modify the header to an invalid CPUID and (apparently) use BIOS-level logic to switch to that microcode under certain situations. Other vendors (e.g. MSI) appear to maintain both 104 and newer microcodes for that CPUID without messing with the header.
