PBO Offset, is lower always better?

Rikk · 28 Nov 2023 at 22:06

This is on a 7950X3d, Asus x670E hero MB.
Sorry if this is a dumb question but I'm new to PBO tuning(my 5800x3d I just set to -30 and that was it..) and have been trying to soak up as much info as I can and can't find an answer to this.

All I'm currently doing is using Ryzen master to lower each core by 5 at a time and using the validate option to stress test it with the assumption that when it fails I can start to wind it back a bit at a time.
BUT, is it always, lower is better?

Joxeon · 28 Nov 2023 at 22:26

Lower is always better providing it is stable as by lowering the curve your essentially undervolting the chip which will allow it to hit the same clocks with less power or even use the extra power/thermal headroom to boost to even higher clocks.

SpudMaster · 29 Nov 2023 at 08:46

My understanding is that lower will allow the cpu to boost higher because you will be supplying less voltage to the cpu and therefore lowers temps which in turn tells the boost algorithm to go higher as you will have greater thermal headroom. Of course if your cpu is unstable then lower isn't better and if your using CO per core them it becomes a balancing act finding the best values for each individual core.

random_matt · 29 Nov 2023 at 16:52

I spent ages on my 7900X3D and eventually just put all cores on -20. Boosts to 5.7 too, I have not tried the additional 100mhz. Additionally, I do not waste my time with crappy software when changing settings within the bios is simply superior.

El Jimben · 29 Nov 2023 at 19:13

Rikk said:
This is on a 7950X3d, Asus x670E hero MB.
Sorry if this is a dumb question but I'm new to PBO tuning(my 5800x3d I just set to -30 and that was it..) and have been trying to soak up as much info as I can and can't find an answer to this.

All I'm currently doing is using Ryzen master to lower each core by 5 at a time and using the validate option to stress test it with the assumption that when it fails I can start to wind it back a bit at a time.
BUT, is it always, lower is better?

It's probably worth having a read of the main X3D thread in the CPU sub-forum but I'm sure that I read in there that the strongest 2 cores should actually have less offset than the rest.

Jamin280672 · 29 Nov 2023 at 20:04

Yes lower is better, not sure id trust Ryzen master validate option to confirm all is fine though, I use Y-Cruncher 1-7-0, if any cores are playing up or too low of a curve.....ie not getting enough voltage to remain stable, Y-Cruncher will tell you exactly which cores it is.

random_matt · 29 Nov 2023 at 22:33

Jamin280672 said:
Yes lower is better, not sure id trust Ryzen master validate option to confirm all is fine though, I use Y-Cruncher 1-7-0, if any cores are playing up or too low of a curve.....ie not getting enough voltage to remain stable, Y-Cruncher will tell you exactly which cores it is.

Y-Cruncher was beyond useless for me. I had -30 on all cores and it passed consistently for hours upon hours. Once you start a game, BSOD, a load of crap to me.

Jamin280672 · 30 Nov 2023 at 13:55

random_matt said:
Y-Cruncher was beyond useless for me. I had -30 on all cores and it passed consistently for hours upon hours. Once you start a game, BSOD, a load of crap to me.

Strange, did you use the component tester ? I've never heard that before, it's pretty much the opposite for me, it passes everything then I run y cruncher and it stops on crashing cores.

It tests everything separately FFT, VST, N64, AVX mem controller etc

PieEater · 30 Nov 2023 at 14:05

I don't think you can say lower is always better, if that were the case AMD would have it set lower in the AGESA settings so everyone benefitted and they could claim more performance from their chips. Whether lower PBO settings are better for you or or not depends on a number of variables but they can be beneficial in allowing cores to boost higher for longer periods due to less voltage / heat being applied. As above it is important to test stability in the apps / games you use and to confirm that performance isn't being degraded rather than improved.

random_matt · 30 Nov 2023 at 14:09

Jamin280672 said:
Strange, did you use the component tester ? I've never heard that before, it's pretty much the opposite for me, it passes everything then I run y cruncher and it stops on crashing cores.

It tests everything separately FFT, VST, N64, AVX mem controller etc

I used the settings listed by LtMatt in one of the threads. I cannot remember what those were, but I am fine with the settings I have now anyway.

Jamin280672 · 30 Nov 2023 at 16:11

PieEater said:
I don't think you can say lower is always better, if that were the case AMD would have it set lower in the AGESA settings so everyone benefitted

Thats not possible as every core on every CCD is slightly different, thats why they make it user controlled, some cores do -0, others -5 and others yet -20 on just one CPU.

PieEater · 30 Nov 2023 at 17:20

Jamin280672 said:
Thats not possible as every core on every CCD is slightly different, thats why they make it user controlled, some cores do -0, others -5 and others yet -20 on just one CPU.

Rikk · 30 Nov 2023 at 21:41

Thanks all, given me something to think about, I'll download y-cruncher as I was looking for something that would tell me which cores have failed

Jamin280672 · 1 Dec 2023 at 04:51

Rikk said:
Thanks all, given me something to think about, I'll download y-cruncher as I was looking for something that would tell me which cores have failed

If you use Y-Cruncher, select options 1-7-0

1 is component tester
7 is all tests
0 is start tests.

PBO Offset, is lower always better?

​