Unsure about building due to (still) worrying SoC voltages (7950X + Strix X670E + 2x32 6000MHz)

[Xeraser]

I've had the parts stated above since February/March except the RAM (a G.Skill 6000 MHz CL30-40-40-96 1.40V kit), by the time it finally got here the whole exploding 7000s debacle had already happened.
My board is the Strix X670E-E rather than the oddly more common(ly explosive) Crosshair Hero and I've found reports of SoC voltage spikes past 1.3 (varying from 1.35 all the way to 1.4+) even on the latest AGESA 1.0.0.7.a BIOS (1416 for this particular board) on both X3Ds and worryingly enough non-X3Ds.

Could other users with the same CPU/MB combo report their SoC voltages? Some claim the spikes only occur with the SoC set to auto, some claim that they occur even with SoC set manually to something as "low" as 1.2v (as of BIOS 1416)
I don't want to play voltage watchdog but I can't return the board either despite having never opened the box and I don't really want to risk frying anything attached to the motherboard. I've been waiting for an AGESA 1.0.0.9 BIOS for two months but I'm very close to giving up.

[Silent_Scone]

Hello,

Please note that the software readings for the Super IO on the Strix read higher than what the CPU receives as the board uses socket sense measurement. The only software sensor you should pay attention to [for VSOC] is CPU VDDCR_SoC which is listed under SVI3 telemetry in HWINFO - anything else can be ignored. The SVI3 is AMD's own internal telemetry and can also be monitored from Ryzen Master.

Please be sure not to run multiple tools polling the system as this can cause polling errors.

This video demonstrates VSOC behaviour when measured correctly from source, and also why other methods will report higher than what the CPU actually receives.

[Xeraser]

Hi, I've seen multiple reports of spikes still happening on the latest BIOS when monitoring VDDCR_SoC with hwinfo, more commonly on B650 boards but I've definitely seen a handful of reports for my exact motherboard. We're talking about spikes of anywhere from "slightly" over 1.3v (1.348v seems to be the most common value) all the way to 1.45v - again, whether the limit is set manually or left on auto. That's what's worrying me. I know the chance of a catastrophic failure is nil now but silicon degradation over time still worries me greatly.

[Silent_Scone]

As I’ve mentioned above, sounds like a polling error. The voltage is capped in firmware. 

The SVI maximum  can be misreported sometimes due to polling overlap. This can occur if other tools are polling the system or if there is a mailbox error.

[Xeraser]

Alright, I'll build it but I'm gonna keep an eye on VDDCR_SoC just in case. I don't think the users that shared their reports were running more than one monitoring software or somehow polling the systems with multiple tools at once but I can't really be sure, the 1.348v figure seems to be really common among those who still have issues so maybe it really is the monitoring software consistently misreporting it. Either way, could you elaborate on the "mailbox error" if possible?

By the way, I'm going to be limiting my 7950X to 125W manually or "automatically" with the built-in 105W ECO mode (either from the BIOS itself or within Ryzen Master in case they added a setting for it), could that affect my system's stability? Would I have to run my RAM at lower than its advertised clock speed/timings?

[Silent_Scone]

Little polling joke in there 😂. When the system checks over API for device readiness the data may not be able to be received or is so incorrectly. 
It’s important to pay attention to the current voltage value for SVI3 readings

The applied VSOC will have an impact on memory stability if it is insufficient for that CPU to operate at a given frequency.

[Xeraser]

I see, thanks. Guess I'll build and have hwinfo log a few benchmark sessions to see if the voltage spikes at all or if it gets some obviously wrong readings. Will update if I find anything wrong