Boglav, DM,
This is by no means an isolated problem! It's not confined to the P55-USB3 and the P55-UD3, because I get this happening with my P55-US3L. The reported CPU temperature starts off as something quite sensible but then after a random period it starts showing figures like +3 degrees C, 0 degrees C, -20 degrees C.
As you know, DM, I have even worse problems with my board, but I nonetheless reported this particular one to GGTS several weeks ago as an aside to the others. They made no comment on it whatever and I've had no feedback from them at all on it.
I myself found that experimenting with the power-reducing settings of C1E, C3/C6/C7, EIST, or turning off the CPU temperature monitoring had absolutely no affect on this. This may differ from board model to board model, though.
So, boglav, you've been somewhat wasting your time, looking for a cause elsewhere. It's pretty obvious that this is a BIOS fault, or at least highly related to the BIOS. It's as plain as the nose on your face. And I guess that what's happened - and this must have taken place a good while ago - is that Gigabyte have taken a core BIOS and applied it across a range of P55s. Thus, due to odd differences in the BIOSs, the temperature-reporting error pops up on only certain of those models.
I get the impression that GGTS are in denial about it. I suspect they find it highly embarrassing, given that they've made a bit of a mess of adapting the core BIOS to individual boards. But the lack of any purely software-based solution from Gigabyte suggests it might be down to a defective temperature sensor that they've used on the board, and perhaps no amount of tweaking of the BIOS can correct it. If so, then for obvious reasons there won't be a solution forthcoming.
The only sensible workaround is to forget about meddling with those power-saving settings and instead fit manually-driven fan controllers in the PC. At least you'll then be assured that the CPU fan will always be running at the revs at which you set it, irrespective of what the BIOS reckons the CPU core temperature is.
Gigabyte may well have taken a cautious and piece-meal approach to initially finding a solution via the BIOS. But even if a BIOS update were to work, Gigabyte would need to be systematic in updating other BIOSs for the other boards that are known to be affected. Thus far, there's no evidence that they're doing this.
BTW, I'm fairly certain I mentioned this happening with my own board several weeks ago in these forums.
You're not alone in this, boglav. You haven't been imagining it!