New BIOS "7E06v1C" for MSI Z790-P WIFI DDR4 VCORE much higher

[jpwisegu153402d9]

Hi, I have an i7-13700K with a MSI Z790-P WIFI DDR4 and this morning I flashed the BIOS to the new "7E06v1C" version containing the new microcode and fixes. Afterwards, I noticed that the VCORE is now much higher than it was with the previous versions. I was using BIOS "7E06v1B."

I have always used default BIOS settings outside of enabling XMP and setting the CPU voltage to an offset of -0.150 in the BIOS. Power settings 253W / 253W and 307A. But after flashing today, then running the CPU-Z stress test with HWINFO64, I noticed where the VCORE with the previous BIOS was around 1.18v - 1.21v under load with temps in the low 70's (very stable) it is now going to 1.34v with temps in the upper 80's (?) All using the same settings (that I am aware of).

I also noticed annoying slowdowns when trying to navigate the BIOS and the voltage dropping further than I have seen.

Anway, flashing back to the previous BIOS version for now. But, does anyone have any idea why this could be? Anyone similar experiences? Any info is much appreciated.

 

[Raxiel]

Check the Lite Load setting, I've seen reports it's 'auto' default is much higher on a lot of the new BIOS files. You can probably drop it quite a bit while maintaining stability.

Note that the 0x125 microcode in this one isn't the 0x129 that supposedly fixes the voltage spikes that Intel asserts is causing the degradation in some chips, so there will hopefully be another BIOS before the end of the month, although I expect that one will be marked as Beta.

Since 0x125 is apparently to fix the TVB bug that (afaik) only affects the i9, I'm going to stick with 7E06v1B myself, with manually imposed power and current limits, and a small undervolt, until the next one is released.

 

[jpwisegu153402d9]

Ah OK. I thought this was the one with those fixes. But I see now it says 0x125 on the page lol. Well, now that I realize that I will just stay with 7E06v1B as well until the 0x129 fixes come.

Thanks for the reply though, much appreciated!

 

[gooddegenerat14e602c9]

I got the 0x129 fixed one for my Z790 Tomahawk WiFi board today and it messes up the voltages and whatnot too. By default it was set to Mode 18 and the PL1/2 is now 181/181w instead of 125/181w. I don't know what MSI is doing but this seemingly just made things worse. Maybe they felt confident that intel fixed the degradation issue and decided to go insane with the voltages again anyway? Lowered my CPU Lite Load back down to Mode 8 instead on my 13600KF. Mode 6 degraded performance a bit in Cinebench and whatnot. Mode 8 seems to maintain it. Haven't tried mode 7 as of yet.

Either way, a regular user who flashes their BIOS isn't going to know why their CPU got higher temps when they update, so I just don't get why MSI did it this way..

 

[Raxiel]

181W/181W (and 200A) is correct for an i5 K processor because they have the performance profile, its only the non K parts that require P1=125W and Tau=56s
That said, at least under the current BIOS and a sensible Lite Load, my i5 can sit indefinitely at the multiplier limits of x51 & x39 on the P and E cores respectively without ever hitting 181W

As for WHY the voltage is so high, I'm no expert but I can see it being something like - the problem wasn't high voltage per-se, but unintended big spikes on top of that high voltage, and now the spikes have been eliminated it's safe to go back to adding a bit more voltage for maximum stability (especially in chips that have already degraded a bit) and let the thermal limit be the governing limit. - not saying I agree, just what I think it is

 

[Raxiel]

Well, I finally found a writeup of the CVE that had a fix bundled up with the 0x125 microcode update and decided to flash it after all.
As expected, the auto lite load was set to 18 or 1.7mohm for AC & DC. That bumped up the voltage around 0.15V compared to what I was running before. It didn't get hot enough to throttle because it ran straight into ICCMax/PL4.
Re set it to Lite load 7, which now gives 0.6mohm for AC/DC, a bit more than before (0.35/0.80) but going to LL6 triggered CEP so I added a-0.03mv offset instead. I did have to bump the ICCMax up a bit. Now its running at the same performance for the same power as before, but only after a lot more fiddling. Perhaps disabling CEP would have made it simpler but I'd rather not do that if I don't have to.