I don't fully agree with that. To me it looks like (and this is also mentioned in some places other than reddit) that there are conflicts between BIOS, Ryzen Master, and maybe some other thing (power plans?).
Also, if I change the same value in one place (Extreme Tweaker > X) and in another place (Advanced > X), then it works fine in one case, and the PC doesn't boot in the other case. Something is definitely wrong here. The specific value that behaves like this for me is Precision Boost Overdrive set to Enabled.
So if I adjust some values, the system is stable, I am getting faster rendering, lower temperatures, and less noisy fans - then why shouldn't I be doing this, and why isn't it working like this out of the box?
Nothing is wrong anywhere :- ). But let's go over it:
- Software measured voltages are highly inaccurate. +/- up to 20perc.
- It's to be noted whether Ryzen Master is used simply as monitoring tool (where is is very crude and limited) or as overriding test setup (or real-time profile switching) for bios settings. This app has multiple purposes, but ultimately isn't needed at all.
- Ryzen Master works better than its other software alternatives (which should be ignored), but even than is inferior option to settings things in bios as it lack all the options the bios offers and that can end up in conflict. Bios offers multiple power regulation settings that influence each other, if only subset is changed in RM it can lead to compatibility issues.
Let's go back to voltages, but it's first good to remind that currently only HWinfo and CPU-Z correctly refers to both VID (Nominal requested voltage from CPU) and V-Core (actual voltage). AMD themselves though suggest CPU-Z to avoid observer effect.
I found HWiNFO to work correctly here as well though with latest updates. Never run multiple monitoring tools to get better accuracy (which includes RM here).
I am getting faster rendering, lower temperatures
Are you sure? So under-volting caused higher turbos? And what are the measurements of before and after?
Undervolting will always (it's physically impossible for it to be opposite) come at expense of single-core performance because the CPU's V-Core will never reach the request VID. This is regulated based on task demands and it will spike up to 1.5V of actual voltage used to reach single-core turbo, the whole system was designed to be this dynamic. Wild swings from 0.2-1.5 are expected and normal.
High idle voltage is not an issue because the current is low as well.
Don't set PBO on if you don't plan on changing the actual power limits. PBO is unfortunately not functioning properly in any auto-state and it might become abandoned (but well meant) concept from AMD.
Here is quote from Robert Hallock from AMD:
Please do not undervolt the chip or set a maximum processor state of 99%. These are ineffective and/or detrimental changes.
Ultimately, you can play with the system as much as you like, it won't achieve any magic. Regardless of what you believe you see. Army of overclockers
can't can be wrong.
If you enjoy the process...that's something else. But I can't get on board seeing suggestions that there is something to optimize become mainstream even among working crowd.