Ok, first of all, thanks a lot to all of you for taking the time to answer!
Indeed, I can probably piece together the parts to rebuild a BIOS file that works, but I have much less time than what I had back then (as in kids), also the Xeon build is working really stable and I don't have it nearby so that I can fix it if something breaks, so I guess the plan for now is to leave it mostly as it is.
The microcode-on-usb looks really interesting, that I will try that because if it stops working for whatever reason, I can just as easily unplug it and go back to business as usual.
It takes a long time on post to show something on screen, and now I'm wondering if it actually has something to do with the secondary BIOS failing to boot and passing the torch to the primary or something like that. Or maybe it's because of the presumably stuck 20x multiplier for the uncore. Whatever is the cause, the raspberry zero working as a keyboard emulator that presses enter if ping fails works extremely reliably, it started out as a hack-on-trial but I see it now as part of the build.
Regarding the microcode, the build is running Server 2016 which includes microcode updates, and using the command "reg query HKEY_LOCAL_MACHINE\HARDWARE\DESCRIPTION\System\Cen tralProcessor\0" I can see that "Update Revision" and "Previous Update Revision" differ, which in theory means that Windows is actually updating the microcode as expected. The values are 1E000000 for the currently running microcode and 14000000 for the boot/previous, I will investigate if it means that it's on the latest available.
A question I have that I'm not qualified to answer is this: let's asume the cause of the long POST and "overclocking failed" error is indeed the microcode, would giving an updated microcode via usb-boot or windows update actually "fix" whatever is faulting, perhaps the theoretically stuck multiplier on the uncore? Are the timings set on POST and microcode-depending, and further changes on the microcode "on the fly" have less chance of fixing things? My gut feeling is that usb-microcode and windows-microcode are less capable and thus unable to fix timings. More stuff to investigate.
Regarding the 4-core vs 6-core differences, my Xeon X5687 is from the fab a 6-core CPU with 2 cores disabled, which is perfectly fine for me because I value more the higher stock multipliers (highest of the "normal" CPUs for LGA1366) while also retaining the full 12MB L3 cache, meaning 50% more L3 for each CPU compared to a 6-core. Anything that applies to 6-core units I expect to apply to my CPU, it's just as Westmere as any other, just with less cores but faster.
Regarding win-raid, it's an amazing site just to read any hardware geek, I am parsing through posts there. Regarding (at)BIOS, I never touched it and never will. Backups of BIOS, I have all of those released for my motherboard stored locally. Worst case, if I screw something really badly I still have my trusty old i7 920, but again, I hope I don't have to get to that point.
Again, thanks a lot!
Originally Posted by Slayer3032
The newest bios for your board has the 0F microcode for 206C2, that's the common microcode revision that has that bug. I'm not sure if that's the microcode the quad cores use but that's the one that the hexcores do.
I patched in all of the latest pre-spectre microcodes into my bios, if you've done it in the past you can probably piece together enough to get you in the right direction out of what I can drag up from a couple years ago when I did it. There was a decent bit of posts about spectre and microcodes around that time, if you dig through the thread a little and search around on win-raid you should find plenty more as well.
First, you need to update your backup bios. Same with anyone else on a Gigabyte board really. If your bios corrupts or fails a check, it's going to wipe your updated bios and overwrite it with the oldest possible bios which absolutely doesn't support Westmere. I think this post details what you need to do.
Ignore any of the "Beta" bios'es. You don't want to touch
(i hope this doesn't tag that poor guy) and I couldn't even convince them to work with CBROM anyways.
You might have to try a few versions of CBROM before you find one that works for your board too. If you want the latest microcodes, I think MC Extractor can get those for you somehow although I think I ripped mine from newer boards, ect. Making your own NCPUCODE.bin isn't too difficult, maybe some of the tools have improved recently.
After your backup bios is updated and you get that all sorted, all you have to do is flash them and if it doesn't work you just revert to your backup bios.