What's new
Carbonite

South Africa's Top Online Tech Classifieds!
Register a free account today to become a member! (No Under 18's)
Home of C.U.D.

For Ryzen 5000 Users - Undervolting = Win

eyetac

Legendary Member
Rating - 100%
82   0   0
Joined
Apr 29, 2020
Messages
1,059
Reaction score
1,143
Points
5,335
Location
Randburg
Hi guys

The latest AGESA update is rolling out and it's great.
I undervolted my CPU as per the video below and my temps have dropped and my performance has gone up!
I dropped around 5 degrees and my all core boost went from 4.3ghz to 4.5ghz.
Witchcraft :D
5900x and 5800x get even better results.

Check it out
(52) Ryzen 5000 Undervolting with PBO2 – Absolutely Worth Doing - YouTube
 
I see the latest version for our boards is a BETA version. Is this the one that your used?

What offset did you use?
 
Will check this out later this evening
I see the latest version for our boards is a BETA version. Is this the one that your used?

What offset did you use?
Latest for my board is F31, no beta options.
 
Yeah, the Beta Bios

Do you mean the Curve Magnitude?
I tried 14, but it failed, so I went with 12 as he recommends for the 5950x in the video.
Stable AF and gained 1500 points in Cinebench R23.
 
Ok so yea, you need PBO2 and Agesa 1.1.8.0 available to get the curve magnitude option, which is not available yet on the Aorus Xtreme X570, latest is Agesa 1.1.0.0D.
So can't even test this new method of undervolting :(
But then again with a Cinebench score of 11346 with a static OC at 4.4ghz it's not so bad...
 
Ok so yea, you need PBO2 and Agesa 1.1.8.0 available to get the curve magnitude option, which is not available yet on the Aorus Xtreme X570, latest is Agesa 1.1.0.0D.
So can't even test this new method of undervolting :(
But then again with a Cinebench score of 11346 with a static OC at 4.4ghz it's not so bad...
Nope. Got the same board as eyetac and even though the last few BIOS updates are all marked as BETA, it at least includes Agesa 1.1.9 (v1602).
 
I see the latest version for our boards is a BETA version. Is this the one that your used?

What offset did you use?

I tried to upgrade to the Beta BIOS for my X570 Tomahawk and the upgrade failed.

Seriously want to get the new AGESA update as it fixes the FCLK overclocking and enables SAM.
 
Well I tried what he has stated in the video as in disable the power limits and my performance tanked but cpu temps were like 20c lower. I leave the power limits which you can set manually or as per motherboard, or leave it AUTO or disable. I leave it AUTO and performance increase is seriously great to say the least. Unfortunately he could not go lower than negative 12 for the vcurve off-set on the 5950X so his numbers not that great. I can go all the way to negative 22 and it is working well but my cpu temps are in the mid to high 80c range. So I started to play around with other settings and I found that I could put a negative vcore off-set of -0.0750 and it dropped the core temps to around 80c or just above when running Cinebench 20R. I can go all the way up to -0.1250vcore off-set but performance drop. All core speed now is around 4475 to 4500mhz which is pretty great.

At present I am running Prime95 with AVX and with the current PBO as explained above and my core speed is 4125mhz using 1.025vcore. Temps just now started to go 82c for the ccd temp. At present I am pretty happy with this setup. I know the cpu is not too bad, I mean under Dry Ice I was able to get 5.25ghz using 1.525vcore which is pretty good.
 
Also based on his scores, on a 5950x I am still better off setting manual clock/voltage as my scores are quite a bit higher than his in cinebench...
 
Also based on his scores, on a 5950x I am still better off setting manual clock/voltage as my scores are quite a bit higher than his in cinebench...
See what I explained above.... my Cinebench 20 score setting PBO is 11560 or there about.
 
Nope. Got the same board as eyetac and even though the last few BIOS updates are all marked as BETA, it at least includes Agesa 1.1.9 (v1602).
I meant on my board the latest is 1.1.0.0D, you need at least Agesa 1.1.8 to get the settings, which is not yet available on the latest bios for the X570 Aorus Xtreme rev1.1.
No new beta bios available yet either, as latest is F31 (stable)
 
My board has BETA drivers available as well. Will try tonight... Didnt have time last night as I was busy streaming... With only 1 viewer... ME... LOL

ROG CROSSHAIR VIII HERO BIOS 3102
Update AMD AM4 AGESA V2 PI 1.1.9.0.
Improve ReSize Bar function
Improve RAID function
Improve system performance and compatibility
 
See what I explained above.... my Cinebench 20 score setting PBO is 11560 or there about.
even with your above, the win for doing this vs static 4.5ghz OC is basically nothing...
temps are in the same range, score in the same range (keep in mind my score was with 4.4Ghz static).
So really no win even if the new bios update comes out for my board with the new Agesa version.
 
even with your above, the win for doing this vs static 4.5ghz OC is basically nothing...
temps are in the same range, score in the same range (keep in mind my score was with 4.4Ghz static).
So really no win even if the new bios update comes out for my board with the new Agesa version.
But do you not then get better single core performance as opposed to locking everything at 4.5ghz?
I have 9 cores that overboost to 5ghz all the time.
 
even with your above, the win for doing this vs static 4.5ghz OC is basically nothing...
temps are in the same range, score in the same range (keep in mind my score was with 4.4Ghz static).
So really no win even if the new bios update comes out for my board with the new Agesa version.
Further on this.
I am sitting with a max cpu temp of around 60-61c at 4.4ghz, so the ~20c further drop in temp for ~600 points less on Cinebench is more than worth it for me.
 
But do you not then get better single core performance as opposed to locking everything at 4.5ghz?
I have 9 cores that overboost to 5ghz all the time.
Probably, but I don't do much single threaded tasks, so single threaded performance doesn't do much for me.
And single threaded performance on tasks on 4.4ghz vs 5.0ghz again is minimal at best.
 
Is there any harm in setting a negative voltage offset of +- 0.048v with pbo2 on ?
 
even with your above, the win for doing this vs static 4.5ghz OC is basically nothing...
temps are in the same range, score in the same range (keep in mind my score was with 4.4Ghz static).
So really no win even if the new bios update comes out for my board with the new Agesa version.
Okay I'll put it to you in perspective. Taken that I have just ran Prime95 small fft with AVX using my PBO settings and my core speed was sitting at 4.125 to 4.15ghz with the vcore hovering around 1.025vcore and temps still sat around 83c. Now with a manual overclock of 4.4ghz you will be using around 1.2 to 1.25vcore to have it stable in Prime95 small fft with AVX which will take your core temps up to the 90c mark.

I am an avid overclocker, I mean my processors has been under Dry Ice or LN2 so many times I cannot remember how many times. I much prefer a static all core overclock, but I found that my results for all core fine tuning PBO is almost the same as with a static overclock and I do not lose any single core performance. It is easy to overclock and run Cinebench but stability can only be certain if you use a proper program o test stability.
 
I got the beta bios for agesa 1.1.9, so I'll give this a go tonight. I'm happy with my manual OC, but based on this video I could get that single core boost on top of my current manual OC.

Will report on me findings.

Currently I hit 9150 on CB20 and 3950 on CB15 with a 4.6ghz all core with 1.27v. Temps on Cinebench is 75 after some runs, on OCCT with medium fft and AVX it sits in the 60s. I don't really even consider small FFT since I will never see that kind of use on my pc, but it will definitely hit 90 then. Small FFT on 1.27v should actually reach 100 if you ask me, it's such an unrealistic load.
 
Okay I'll put it to you in perspective. Taken that I have just ran Prime95 small fft with AVX using my PBO settings and my core speed was sitting at 4.125 to 4.15ghz with the vcore hovering around 1.025vcore and temps still sat around 83c. Now with a manual overclock of 4.4ghz you will be using around 1.2 to 1.25vcore to have it stable in Prime95 small fft with AVX which will take your core temps up to the 90c mark.
Well I decided to run P95 (not something I run often).
Set it to Small fft with AVX.
Max temp of 84.1c with cpu core voltage at 1.144v
Gotta love AMD and their voltage droop...

So still much of a muchness when it comes to 5950x...

I agree that the PBO2 method (if available) is awesome for the 5900x and lower where the possible boost is quite high, but for the 5950x from what I can see, it's barely worth the effort of a. waiting for a beta bios that includes the updates, and then b. digging around to find the settings which is hidden all over the place.
 
Well I decided to run P95 (not something I run often).
Set it to Small fft with AVX.
Max temp of 84.1c with cpu core voltage at 1.144v
Gotta love AMD and their voltage droop...

So still much of a muchness when it comes to 5950x...

I agree that the PBO2 method (if available) is awesome for the 5900x and lower where the possible boost is quite high, but for the 5950x from what I can see, it's barely worth the effort of a. waiting for a beta bios that includes the updates, and then b. digging around to find the settings which is hidden all over the place.
Well if you running Prime95 small fft at 4.4ghz with vcore at 1.144vcore after droop then it is great to say the least. Mine would not do 4.4ghz using that vcore or vcore that low.

The nice thing about PBO 2.0 with the vcurve optimizer is that you can get all core speeds around 4.45 to 4.5ghz using it. Under Prime 95 small fft with AVX the all core speed would be around 4.15ghz with a much lower vcore but it would be perfectly stable.
 
Pbo 2 still not working lekker on the tuf board, but manually it does well. I'll revisit with a stable bios later.

I do have a pretty good cpu it seems though, I can do 4.8ghz on the 1st ccx and 4.6ghz on the 2nd ccx with only 1.3v stable. Also 4.7ghz all core 1.3v stable.

But I just put it back on 4.6ghz all core, it adds like 7 celsius when compared to 1.27v 4.6ghz which I feel is not worth the performance increase in games (which is probably less than 0.5fps on 4k)
 
Last edited:
Well if you running Prime95 small fft at 4.4ghz with vcore at 1.144vcore after droop then it is great to say the least. Mine would not do 4.4ghz using that vcore or vcore that low.

The nice thing about PBO 2.0 with the vcurve optimizer is that you can get all core speeds around 4.45 to 4.5ghz using it. Under Prime 95 small fft with AVX the all core speed would be around 4.15ghz with a much lower vcore but it would be perfectly stable.
So I actually have the PBO 2.0 settings in my bios aparantly, but not sure if it's working as intended (agesa version requirement and what not).
Set a negative curve of 20 and boot temps were like 15-20c higher than on static OC (Boot temp was 75c, idle 60c / Static Boot temp is 60c and idle is 40-45c).
turbo barely ran 4.1Ghz on all cores
And Cinebench R20 scored 150 points higher than stock...

Either I am doing something wrong, or shits not working.
Back to static OC I go for now.
 
So I was pleasantly surprised at how much my ram could be tweaked, on the Z390 I couldn't even touch my ram and it wouldn't even post.

I got around 75% with optimizing my timings, will continue tonight.

So far I'm stable at 3600mhz, 15,18,18,19,36. Will do further testing tonight!

That's down from 3600mhz, 16,19,19,19,39 so I'm already happy with the results.
 
So I actually have the PBO 2.0 settings in my bios aparantly, but not sure if it's working as intended (agesa version requirement and what not).
Set a negative curve of 20 and boot temps were like 15-20c higher than on static OC (Boot temp was 75c, idle 60c / Static Boot temp is 60c and idle is 40-45c).
turbo barely ran 4.1Ghz on all cores
And Cinebench R20 scored 150 points higher than stock...

Either I am doing something wrong, or shits not working.
Back to static OC I go for now.
Okay first of all, the limits for PPT PTC and EDC or whatever it is called, in the video states set it to disable. I found that this completely tanked performance. I leave that on AUTO. It does mean 15 to 20c higher core temps but performance is way higher than with those limits disabled. Negative 20 is good if it works for you. If you do test it again, you will see all core clocks around 4475 to 4500mhz. You can decrease the core temps by setting a negative offset for the vcore, like -0.0750.... I can go all the way to -0.1250 but performance decrease a lot when the negative off-set is too low.

I ran Cinebench 2 this morning with the limits disabled and my ccd temps did not even go above 59c but performance was 10426 instead of 11500+. At present I will leave it as is as I prefer the lower temps unless I bench the system.
 
So I've run into problems with this after flashing a new beta bios yesterday.
My max all core boost locks to 3.7ghz and won't budge higher.
Single core is fine and hits 5ghz when boosting.
Put everything back to stock with PBO2 enabled and getting into 28k's in cinebench R23 again, but smashing into the thermal wall of 90 degrees and the cpu throttling to 4.2 ghz.. Boo. I was maxing out at 80 degrees or so and all core at 4.5ghz before loading this new 1802 bios.

Hoping they release an updated bios soon.
 
I don't think the Tuf X570 knows what to do with the new voltage curve settings.

At standard PBO 2 settings it works fine, hitting 4.4ghz all core, 4.95ghz single core. But it hits like 86-87c just by running cinebench a few times because it hits 1.39v all core.

If I put 10 negative ticks on the voltage curve I get the same voltage but higher speed, so basically still 1.39v but now I get 4.5ghz. At 20 ticks I still get 1.39v but now I get 4.6ghz all core. And more than those ticks crashes the moment you run a bench.

Also if I disable the limits as mentioned then the curve does even less, now it still boosts the same speed with the elevated vcore at 1.39v, and now it crashes even with 15 negative ticks. It literally doesn't do what it should which is lower the voltage.

So yeah that feature is obviously not lekker on this mobo rofl, but manually I get great results so it's not a bother. I like my 4.6ghz all core boost at 1.27v, the temps are lekker in the ball park and it's just so stable.

Cinebench hits 75c then, which is a whole 12c less for the same all core boost. I would give up single core performance for that any day.

In anycase my cpu is stable on 4.8ghz ccd1 and 4.6ghz ccd2 on 1.30v. Then I hit 81c on cinebench after a few runs which is still better than manual pbo 2. But I prefer to be below 80c so 4.6ghz it is.

Regardless of the settings it struggles to pass 60c in games, but alas I'm full of shit that way hehe. At 4.8ghz/4.6ghz 1.30v it idles around 34 and it hits average of 65c on OCCT with medium FFT AVX2, so I guess it's a very safe temp. I'll think more on it. I have seen that it boosts single core performance well and since it's 12 threads running at 4.8ghz it's quite perfect for games.

This was tested on both the agesa 1.1.9.0 and 1.1.2.0 bios so I don't see this improving anytime soon.
 
Last edited:
@VPII

What should this voltage be? See the image below. It's not the vcore. I set mine capped at 1.1v and I'm still rock solid on 1.27v 4.6ghz, and this caused my temps to drop a bit as well because previously I used to set this manually to match the vcore @ 1.27v.

Initially I thought this was vcore but its not, the vcore can be seen on HWMonitor. What should this core voltage be when manual overclocking?

Note : This picture is not mine I just googled it since I'm at work.

For some reason my Cinebench score also increased to almost 9200, which is like almost 50 more (but perhaps this is because of the agesa bios 1.1.2.0 update so not sure what caused the increase)
 

Attachments

  • EPZrV5lW4AEwuJZ.jpg
    EPZrV5lW4AEwuJZ.jpg
    309.6 KB · Views: 21
Last edited:
Perhaps that's the reason I thought PBO 2 wasn't working correctly, because that is the voltage that was hitting 1.39v. I never even checked what the vcore was with PBO 2 in hwmonitor.

Perhaps I must revisit it tonight for interest sakes.
 

Users who are viewing this thread

Back
Top Bottom