Important reminder about voltages for overclockers

Gamersnexus.net recently did a bit about safe voltages, which you can find here:

https://www.gamersnexus.net/guides/3251-raven-ridge-soc-voltage-guidelines-how-to-kill-cpu-with-safe-voltage

They're concentrating on Raven Ridge overclocking, but this should apply to pretty much all CPU overclocking: The voltage you set in BIOS or Ryzen Master may not be what the CPU actually receives. The voltage that is shown in HWMonitor and the like may also not be what the CPU actually receives. There are cases where the actual voltage is notably higher than the set or reported voltage, which means that if you are trying to stay just below what AMD declares it safe, you may in reality subjecting the CPU to voltages that will cause degradation over time.

The difference between the set and actual voltage will depend on the motherboard, possible LLC settings etc., so there's no easy way to adjust your settings correctly, if you are aiming for a "safe" value. Unfortunately it is not always clear to which voltages people are referring when talking about these "safe" voltages. Is AMD only talking about what is set in BIOS and if so, have they accounted for a worst case scenario between the set and measured voltage? What about if the safe voltage value is given by a motherboard manufacturer

In short, don't trust voltage values blindly, whether they are claimed safe voltages or voltages reported by software. If in doubt, play it safe. Also, if someone has done some measurements themselves, it would be interesting to know what sort of values ASRock's motherboards are posting. All I know is that according to HWMonitor my AB350M Pro4 uses a ~0.03V higher Vcore (under Prime95 testing) than what I've set in the BIOS. That alone is enough to encourage me to not use a "safe" value of 1.35V in the BIOS.

You'd probably need a multimeter to get more accurate information about the voltages. Using just software to measure the voltages, I don't think you can do much better than HWMonitor. Probably any other quality software will show you the same values as HWMonitor, but you can of course test different options and see what you get.

The voltage you set in BIOS is a target voltage and it's normal that the actual voltage varies based on CPU load, for example. On my system the reported voltage is in practice always higher than what I've set in the BIOS. This means I'm above my target Vcore even when Vdroop kicks in under heavy load. The downside is that the CPU is probably receiving more voltage than what it would need to be stable when it's not under heavy load.

I don't use P-states myself, but I'd guess that's where the issue is. It could be that some program running in the background causes enough CPU load to prevent dropping to lower P-states. Otherwise those numbers seem normal. You've set 1.375V in BIOS and you're seeing 1.375V normally, then under heavy load Vdroop takes that voltage to 1.337V. I think this is pretty much what you should see without LLC.

Hmm. Maybe I was wrond and that's what you should expect with linear LLC that's not overly agressive. After all, the drop is not huge - at least on older systems the Vdroop could be over 0.1V without LLC. Have you checked what numbers you get with different LLC settings?

After some research i found the answer to my question : https://youtu.be/2DGDViPRh9k so i leave it here incase that can help someone .