News AMD Introduces Precision Boost Overdrive 2, Boosts Single Thread Tremendously

spongiemaster

Admirable
Dec 12, 2019
2,273
1,277
7,560
AMD introduces its new Precision Boost Overdrive 2 tech that boosts the overclocking capability of Ryzen 5000 series processors.

AMD Introduces Precision Boost Overdrive 2, Boosts Single Thread Tremendously : Read more
Are there other benchmarks that have been revealed besides the Cinebench results here? 5800x went from 628 to 638. That's a 1.6% improvement. 5900x improved by 1.9%. You call that boosting single thread "tremendously?" You have tremendously low standards if less than 2% qualifies for that label.
 
  • Like
Reactions: thisisaname

saunupe1911

Distinguished
Apr 17, 2016
203
74
18,660
PBO overclocking is very confusing. I have a 5900x and Gigabyte X570 Aorus Master. I don't see any performance benefits except higher voltage with PBO turned on. Cinebench and gaming performance was basically equal to default settings as you mentioned in your graph.

I kept it simple and set the CPU clock to 4400Mhz and voltage to 1.20. This all core overclock is much more efficient while actual voltage rarely goes over 1.14. PBO had it going up to 1.3.
 

InvalidError

Titan
Moderator
What an ingenious scheme. You convince your customers to try a utility that provides a "tremendous" performance increase of less than 2% and they will voluntarily void their warranty. I bet someone in the support group came up with this idea.
Got to appreciate the dichotomy of not condoning overclocking on one hand while promoting the benefits and the tools facilitating voiding your warranty on the other.
 

waltc3

Reputable
Aug 4, 2019
420
223
5,060
I'm running PBO on my 3900X with the latest 1.1.0.0 AGESA--well, been running it for the last 12 months at least (x570 Aorus Master)--and I routinely get 4.65GHz max single-core boost--air-cooled, stock fan, stock clocks, on 3-5 cores (according to HWinfo). From what I've read the main difference between Zen2 PBO and Zen3 PBO is not just a bit higher max-single-thread boost clocks (which are lost if you manually overclock all the cores as boost is disabled) but 5000-series PBOs are also supposed to deliver the higher boost in a more sustained manner--at least "sustained" in comparison with Zen2 PBO boost clocks, that is. Yes, the Zen 2 max boost is there but only in lightly threaded workloads, and very briefly, too.

It's fun to tinker with, but the differences are 200-300MHz, max, over what the CPU will do boosting or overclocked on all cores. Sometimes, people split hairs over a mere 50Mhz difference in max boost clocks. The single-threaded Cinebench tests are, I think, rather silly, as no one is going to render professionally using this program (or anything like it) in single-thread mode...;) Defeats the entire purpose of multicore SMT CPUs, imo!
 

Makaveli

Splendid
What an ingenious scheme. You convince your customers to try a utility that provides a "tremendous" performance increase of less than 2% and they will voluntarily void their warranty. I bet someone in the support group came up with this idea.

When was the last time you had to RMA a processor i've been doing this 20+ years and never had to once. It will be fine aslong as your smart enough to be using proper cooling and not stock POS cooler.
 
  • Like
Reactions: Thunder64

CerianK

Distinguished
Nov 7, 2008
260
50
18,870
Based on some of the comments I might assume that not everyone is familiar with the intricacies of under-volting, which is part of the equation with PBO2. I know someone that used it to good effect on Zen, so perhaps Zen 3 will benefit from it as well. But because of the changes in Zen 2/3, it must be dynamically modulated to get a benefit (for Zen 3). Just a guess, but I would wait for independent reviews on PBO2 / new AGESA before passing judgement.

Also, I understood the new AGESA would be required to assist with DDR4-4000 1:1 stability, for those marginally capable, but could not do it on the current AGESA. Just more knobs to twist and levers to pull, which is all good.

And yes, < 2% performance improvement is hardly noteworthy, which is why I found the 3000XT variants questionable, but those were a different SKU/price, which is not the case here.
 

InvalidError

Titan
Moderator
If the risk for damage is so low, why does it void the warranty?
Likely because it pushes the chip beyond AMD's own confidence interval and it does not want to deal with the increased RMAs from doing so. It probably also does not want to have to deal with people complaining about their chip not PBO/PBO2-ing as high as it used to. Not offering warranty for PBO(2) means you can't bring RMA complaints for anything above stock.
 

deesider

Honorable
Jun 15, 2017
298
135
10,890
Likely because it pushes the chip beyond AMD's own confidence interval and it does not want to deal with the increased RMAs from doing so. It probably also does not want to have to deal with people complaining about their chip not PBO/PBO2-ing as high as it used to. Not offering warranty for PBO(2) means you can't bring RMA complaints for anything above stock.
I suspect their proviso wouldn't apply in countries with strong consumer protection laws, such as EU, Australia, New Zealand etc. In those jurisdictions the actual wording of any warranty is effectively ignored, since there is an obligation under law to provide a warranty, the terms of which are based on the type and value of the goods.

In the case of PBO it's certainly arguable that use of a feature provided by the manufacturer could never void a legally obligated warranty.

- I also note that a proviso on overclocking does not appear in the actual warranty statement; only that an adequate cooler must be used: https://www.amd.com/en/support/kb/warranty-information/rma-03
 

InvalidError

Titan
Moderator
I suspect their proviso wouldn't apply in countries with strong consumer protection laws, such as EU, Australia, New Zealand etc. In those jurisdictions the actual wording of any warranty is effectively ignored, since there is an obligation under law to provide a warranty, the terms of which are based on the type and value of the goods.
I doubt that any such warranty obligations cover destruction by user error.
 

everettfsargent

Honorable
Oct 13, 2017
130
35
10,610
Other than the obvious click bait, it is nice to know that "Tremendously" has a quantitative number attached to it, like ~2.0 percent faster. Next up for TH? Quantitative definitions for gigantic, humongous, enormous and bigly yuge!
 
Last edited:

everettfsargent

Honorable
Oct 13, 2017
130
35
10,610
The 1st thing I will do after building a 5950X system? Take the default three year warranty and get a zero year warranty! Now that would be a "Tremendously" stupid thing to do with an $800US CPU. All for ~2% faster? Yeah, that's the ticket! :(