[SOLVED] What is the best safe PBO scalar for 3700x?

[LMSNK]

I need to know what is the safest pbo that will generally just work. I’m running 4.3ghz at 1.42v

 

[CountMike]

I need to know what is the safest pbo that will generally just work. I’m running 4.3ghz at 1.42v

PBO scalar has influence only on how long top boost lasts before it throttles down from it, has no influence on top frequency.

 

[LMSNK]

PBO scalar has influence only on how long top boost lasts before it throttles down from it, has no influence on top frequency.

I actually know very little about overclocking so I set everything to auto and limit controlled by Mobo, but I need to know what is a safe number for extra performance. I’ve seen people say that only 10x have them better performance. But not sure if it’s safe or worth it.

 

[CountMike]

I actually know very little about overclocking so I set everything to auto and limit controlled by Mobo, but I need to know what is a safe number for extra performance. I’ve seen people say that only 10x have them better performance. But not sure if it’s safe or worth it.

Auto or 10x is not visible in normal work, not even for short benchmarks. Length of boost is determined by core temperature at that core.

 

[drea.drechsler]

Auto or 10x is not visible in normal work, not even for short benchmarks...

I don't agree...but I would agree it will depend a lot on unique system pecularities and how it's set up. On my system I have to enable PBO (take it out of DISABLED) to expose the setting, with PBO in AUTO it does basically nothing whatever scalar value I choose.

But on my system, and with it in MANUAL, using my preferred TDC, EDC and PPT settings, MT performance will increase with scalar. The best CB20 MT scores are at 10X scalar, by about 5-10 points over scalar 5x with several runs averaged. That's not much, but it is measurable in the right benchmark so it can be said to be 'noticeable'. The difference is even greater vs. SCALAR in AUTO but I didn't make the comparison tests.

Also, during thread light single thread boosting (PCMark10) the core voltage is hitting as high 1.49V on 10X but staying under 1.47V for a scalar of 5X. In that kind of BM small variations are hard to find without multiple runs, and it takes a lot of time each run, so it just wasn't worth it. But in the end that extra voltage is something that matters, even though it's expected according to AMD. That's why I settled on 5X scalar for my system.

EDIT: just to clarify, this is on a 3700X system. My understanding is scalar settings can have much greater influence on 2000 series CPU's.