• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

An Excellent Video Guide on How to Configure PBO/Curve Optimizer

Soldato
Joined
29 May 2005
Posts
4,899
Just want to supplement this video with this Reddit post.

I am following this guys methods. It is so far giving me no issues. All my cores are able to boost to 5.025GHz and currently testing my CO values per core and per thread basis using prime

https://www.reddit.com/r/Amd/commen..._source=amp&utm_medium=&utm_content=post_body

I am currently on -20/-10/-25/-25/-25/-25/-25/-25 and every thread is stable with the affinity load with light prime and Aida memory stress test.

I wonder if I should go further on my values. Core 1 and core 2 are my second best and best core respectively.
 
Last edited:
Associate
Joined
20 Jun 2016
Posts
1,308
I must have got lucky with my 5090x my CO is -10 on my best 4 cores and -20 on the rest, my system has been on for 130 hours straight now mining
rMpnVK8.jpg

Are you mining on the CPU too? Then it wont show much instability as you are only working the card.

Stick CPU and card on mine and then do 130hrs........

see if that is stable:cool::D:D:D

#goodluck
 
Associate
Joined
20 Jun 2016
Posts
1,308
Yeah if you are happy awesome.

Nice YouTube guide this btw. There are some 5-series CPU that are absolute lemons and can hardly do any negative curve, even on stock.
But I guess that goes for any silicon. They dont call it silicon lottery for nothing.
 
Caporegime
Joined
17 Mar 2012
Posts
47,578
Location
ARC-L1, Stanton System
I have been using CoreCycler to fine tune mine.

To start set a desired boost clock override and a global negative curve, run CoreCycler, it will tell you which cores error out, if they are all fine set a higher negative curve until one or more cores error.

Make a note of which cores error and go back in the BIOS, set per core negative curve, set all cores that passed to the curve value you had and the ones that error to the value where they passed, then test the remaining cores with a higher value negative curve. Keep doing that until you have all cores individually set.

https://www.overclock.net/threads/corecycler-tool-for-testing-curve-optimizer-settings.1777398/
 
Associate
Joined
7 Dec 2002
Posts
1,833
Location
Frauenfeld, Switzerland
I have been using CoreCycler to fine tune mine.

To start set a desired boost clock override and a global negative curve, run CoreCycler, it will tell you which cores error out, if they are all fine set a higher negative curve until one or more cores error.

Make a note of which cores error and go back in the BIOS, set per core negative curve, set all cores that passed to the curve value you had and the ones that error to the value where they passed, then test the remaining cores with a higher value negative curve. Keep doing that until you have all cores individually set.

https://www.overclock.net/threads/corecycler-tool-for-testing-curve-optimizer-settings.1777398/

Thanks man. I have started using CoreCycler instead of OCCT now. The interface is less modern but it did find an error running overnight so I reduced to a smaller offset on that core. At present I am testing at stock but I might switch to a +50mHz PBO Overdrive offset and retest afterwards. I am not sure it will make much of a difference though as I hardly saw a benefit at +200mHz (which was unstable).
 
Caporegime
Joined
17 Mar 2012
Posts
47,578
Location
ARC-L1, Stanton System
Thanks man. I have started using CoreCycler instead of OCCT now. The interface is less modern but it did find an error running overnight so I reduced to a smaller offset on that core. At present I am testing at stock but I might switch to a +50mHz PBO Overdrive offset and retest afterwards. I am not sure it will make much of a difference though as I hardly saw a benefit at +200mHz (which was unstable).

For high clocks heat is your enemy, as soon as the CPU gets over 65c it starts reducing clocks gradually, if you can keep it under 65c in games it will boost to 5GHz+ on all cores if you have the boost override set as such, for hard MT workloads its impossible to keep them under 65c but the cooler you can keep it the higher it will clock.

Applying a negative cure will reduce temps, sometimes to get higher clocks set a less aggressive boost override and more aggressive negative cure.

Its about finding the right balance, that is also very different from one CPU to the next.
 
Caporegime
Joined
17 Mar 2012
Posts
47,578
Location
ARC-L1, Stanton System
Aslo, don't set higher PTT, TDC and EDC values on a 5800X, there is no reason a 5800X would use more than its stock values, for example PPT is 142 watt, that's the socket limit, if you need to set higher values than that for a 5800X something is set wrong, even at 4.9Ghz full bore R23 it shouldn't be pulling more power than that.

Something like a 5950X is different, effectively two 5800X's, one pulls 110 to 130 watts at 4.8Ghz in R23, double that but minus about 15 to 20 watts for the IO die for all core 4.8Ghz in R23 5950X, so around 220 watts. that's how you get that juicy 32K+ in R23. (2X a 5800X)
 
Last edited:
Associate
Joined
7 Dec 2002
Posts
1,833
Location
Frauenfeld, Switzerland
Aslo, don't set higher PTT, TDC and EDC values on a 5800X, there is no reason a 5800X would use more than its stock values, for example PPT is 142 watt, that's the socket limit, if you need to set higher values than that for a 5800X something is set wrong, even at 4.9Ghz full bore R23 it shouldn't be pulling more power than that.

Something like a 5950X is different, effectively two 5800X's, one pulls 110 to 130 watts at 4.8Ghz in R23, double that but minus about 15 to 20 watts for the IO die for all core 4.8Ghz in R23 5950X, so around 220 watts. that's how you get that juicy 32K+ in R23. (2X a 5800X)

Yep, very true. So I should keep focusing on the curve offsets for now in pursuit of finding the sweet spot on the efficiency curve. I am just under 2000 in the single thread score of Cinebench with my -10 offset on the preferred 2 cores, PPT 110W, TDC 75A, EDC 100A without any boost override which is better than with the PBO default values of 140W etc. The other cores seem stable at -15 for now.

I get what you are saying about temps, but I like my PC staying quiet, so the fans only run at 75% unless it hits 80c. If I just wanted high scores I could probably squeeze a bit more but it would be distracting.

I am going to focus on finding the best curve offsets first to try for the highest stable sustained boosts, then compare what I gain adding in a Boost offset in some gaming benchmarks. If performance is only 1-2 fps more I will probably stick with whatever is cooler.
 

TNA

TNA

Caporegime
Joined
13 Mar 2008
Posts
27,524
Location
Greater London
I’d love to get my hands on your samples. :p
Bet you would… :cry:

To be honest I think in his case it is just pebcak :p

I have been so happy with my 5900X at stock I have have turned off PBO. Might give this a go one day when I fancy some overclocking.

As it is I am more than happy with this CPU and probably won’t bother upgrading for a very long time. Once they can double the performance of each core, I will consider it :D
 
Associate
Joined
22 May 2015
Posts
1,954
Location
Manchester
At stock, my 5950x is getting scores of 23,000 on r23 without PBO. Is this normal? Ram is 4x8gb 3600mhz CAS 16-18-18-36. (multi core that is obviously)

Can be better. This 5900x scores 22,000 with tweaking so I'd look into PBO and CO.

This one is stable at - 30 on all cores with Curve Optimiser and 155a EDC. Boosts to almost 5Ghz in games while staying fairly cool, the extra performance is noticeable.

Watch the video posted in the first post here. I didn't use that particular one but it's not so difficult.

Edit: it IS normal without tweaking.
 

TNA

TNA

Caporegime
Joined
13 Mar 2008
Posts
27,524
Location
Greater London
Can be better. This 5900x scores 22,000 with tweaking so I'd look into PBO and CO.

This one is stable at - 30 on all cores with Curve Optimiser at and 155a EDC. Boosts to almost 5Ghz in games while staying fairly cool, the extra performance is noticeable.

Watch the video posted in the first post here. I didn't use that particular one but it's not so difficult.

Edit: it IS normal without tweaking.

What method did you use? I might have another go at it, just don’t fancy spending ages. What is 155a EDC?
 
Back
Top Bottom