PBO off values are lower than the 90/140, so you'll get a performance boost manually setting those vs stock.
EDC, TDC and PPT do NOT change voltages, therefore they never affect stability - the only possibly setting there that has a stability affect is the +200 setting and undervolting, because you can have them set in a way it rarely or never reaches that +200 to expose its instability. Not every CPU can handle the positive clock boost, and honestly it rarely activates and for such a short period of time it's not worth it.
PBO (without the EDC bug) doesnt cause clock stretching, thats a side effect of undervolting. The CPU's clocks are a result of VID - CPU gets told it's receiving 1.30v VID, so it sets the matching clock speed. Undervolting offsets then send 1.25v instead of 1.30v, to save power and heat.
If the undervolting goes too far, the CPU thinks oops the VID is lowered, time to clock down further - resulting in stretching. (The settings in the BIOS can control this, static voltages for example do this and reduce the max boosts, while offsets and curve optimiser try to prevent the downclocks)
The x3D seems to behave a little differently here and recognises that it's undervolted and downclocks, instead of the effective clocks revealing the difference. I can use -30 and -0.60v offset together, but any lower and it downclocks at load.
true, true, but as soon as pbo with boost override is engaged there may be a possibility that the cpu becomes "undervolted" and clock stretched depending on the chip quality.
i may misunderstand this or not use the right terms but let me try to explain.
e.g. 5700x has a cpu boost max of 4650mhz (fmax), precision boost clock limit is 4850mhz (pbo max).
max. core vid is applied to 4650mhz and can then be stretched up to 4850mhz with pbo boost override.
considered the max. vid is applied to fmax then every clock above that will be stretching the max. vid to a higher clock / pbo max.
the boost override range 0-200 is the range where amd feels like clock stretching is non-existent or reduced to a level where it may not be notice- or measureable.
thats why xfr, pbo is considered overclocking already and will void your warranty.
while the cpu is working with pbo disabled, its ensured by apb and the max. vid. there will be virtually no clock stretching up to 4650mhz: this is the range deemed by amd to exclude clock stretching completely.
pbo disabled or pbo without boost override may ensure clock stretching is completely excluded or at least reduced to a level where its not noticeable or not measureable.
edc and tdc supply amps and theoretically there shouldnt be an effect on voltage but there is with the bug - and there may be more and other effects that havent been discovered yet:
"Going just 1A above the default EDC limit triggers the bug, and it looks like it reduces the voltage by about 75-100mV for both single core and multi core loads."
reducing my edc from say 140 to 90 gave me increased max. corevid, cpu min and everage temps were higher, performance was the same or even better but i believe hell let loose crashed a few times because of that. i didnt have any crashes with edc @ 140 it seemed to me it came from the change.
edc and tdc may be disproportionate to the extent that ppt cant be satured; for max boost override of 200mhz you would also need edc, tdc and ppt maxed out.
its true that they theoretically shouldnt cause any instabilities and are unable to do so.
in each case most of all im limited by an air cooler (lc power cosmo cool lc cc 120), but im pretty satisfied with the results, yet with water im sure i would get higher all core clocks no problem and there may not be that much to get out of a 5700x anyway.
if i remember correctly ryzen 3000 could change frequency and switch cores on or off 1000 times per second or something.
so yes, one could argue pbo max with 200 override doesnt really matter, because its rarely reached or sustained and if then only for very short times.
on the other hand this is exactly what makes out ryzen and maybe that will give you higher performance in some but not in all cases.
im mainly concerned with performance in games.
there are 4 ways of overclocking ryzen:
1.pbo
2.multiplier
3.frequency and voltage
4.custom p0state (frequency and vid)
when i run cinebench e.g. all cores will be maxed out to run at max. possible frequency and voltage which adapt to the temperature, e.g. 4500mhz all core boost @86°C.
but when i run games cores are not maxed out in the same manner, neither all cores run at max. frequency nor 1, 2 or 4 cores run at max. frequency.
actually when i run a game id like to see at least one core running at full speed all the time; what i want to see is that the cpu would run at maximum performance, voltage and temperature similar to the behavior in cinebench. but as far as i understand these are issues of the game / app itself and second the scheduler which is either supplied by software (windows, ctr / hydra or process lasso) or hardware:
cppc auto = windows choses cores instead of algorithm
cppc enabled = hardware algorithm choses cores
for ryzen 5000 amd recommends auto for cppc settings. i have them set to enabled. i wonder what the difference will be if any.
cppc enabled vs disabled
what would i need to do to achieve that? i know that the scheduler of ryzen clock tuner did just that (run fastest core at manually specified clock speed and voltage) but its not compatible with the 5700x. so now i have to find another way and i dont really want to use hydra. with the free version of hydra i have no option to edit the profiles. in ctr i could simply edit the profile and specify a max frequency and max voltage and thats it. now what are the exact methods of ctr? did it set a custom pstate0 or did it set a frequency and voltage? and what did ctr do to cause the fastest core to constantly run at max specified frequency?
what i also do not understand is why there are two overclocking menus in the bios, settings are doubly present and whats the difference between them: amd cbs and amd overclocking.
overall id say it makes more sense to "clock compress" ryzen to its sweet spot as it was done in the benches of ac valhalla ive shared before and this one:
this way you will have reduced clocks, power consumption and thermals but still the same or even better performance. and thats what indicates to me that between pbo-500/ -24, dvo -131mv and pbo+100/ -19 there is "clock stretching" going on; i.e. higher clocks, higher temps, higher power wont yield you much better performance and youre better of going the other way and compressing / condensing it down, i.e. undervolting and of course there may be other and further reasons and explanations for this.
the next "best idea" for overclocking to me seems to disable pbo and go with a custom pstate0 see how high your clocks can go and be stable with a safe max vcore something like 1.36v, 1.38v, disable cppc and cool&quiet. so far i havent tested manual overclocking (frequency and voltage) or custom pstate0 (frequency and vid) yet and what are the differences between them, so im not sure which one is better for me, but i certainly wont do any overclocking by the multiplier (all cores run at max speed all the time).
