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.

AMD Ryzen 5000 and PBO

VPII

Legendary Member
Rating - 100%
39   0   0
Joined
Dec 10, 2010
Messages
1,979
Reaction score
714
Points
5,965
Okay I figured there may be some questions or difficulties getting the most out of your AMD Ryzen 5000 processor using PBO. Now I will be honest, the entire thing of using PBO never really sunk in with me and I was more a fan of using good old manual overclocking, that was until the AMD Ryzen 5000 landed for now I get almost the same all-core overclock speed when running PBO.

Please understand, I am no expert I am only sharing what I found to be working for me.

But how do I do that. Well firstly you motherboard bios need to be flashed with the latest bios with Agesa 1.1.8.0. Most of the X570 motherboard should have this right now and even some X400 boards from what I have gathered.

Now please understand what I am explaining here is not perfect, but it worked for me which is why I want to share it.

Now my implementation may mean higher core temps, I mean into the 80c range but it should be fine but I’ll explain to you how you can swop between the ultimate performance and cooler running temps with one change in the bios.

Firstly you go to control panel > hardware and sound > Power Options. Check in there if you have the Ultimate Performance Plan and if you do set it to that. Now I was surprised when I did not have it so I looked around how to get it and this is how you can get it. Open command prompt as administrator and type: powercfg -duplicatescheme e9a42b02-d5df-448d-aa00-03f14749eb61 and hit enter, it should tell you that the Ultimate Power Plan has been loaded. Now you can go back into control panel same area and select the Ultimate Performance Plan.

Now to restart and enter your bios.

First thing to do in the bios is to go to where you set the vcore (core voltage) for the processor. In the setting should be an option for off-set and upon selecting it, it should open two option, one to select – or + and you select – and the next section you select how much you want to decrease the vcore by. Usually start with 0.050v if this works you can later go down to 0.0625, 0.0750, 0.0875 but the moment you will get to 0.100v your performance will start dropping. This will help reduce core temps under full load.

Now you will need to find the PBO settings, or precision boost overdrive. Your bios may have a search function to locate it but it is possible in the advance amd settings. Now here I will explain the two settings to change between the best performance or reduced performance but with way lower core temps.

Firstly is the PBO limits you can usually leave it auto, manual or disable sometimes even motherboard. Now I have set mine to motherboard as it works, but Auto will do the same. So for performance you leave it Auto or set it to motherboard. If you want to reduce the core temps while also losing all core performance you disable this. Core temp drops will be around 15 to 20c so it is good but at the expense of performance.

Secondly: Precision Boost Overdrive Scalar – Auto as I do not really see the point as yet but will add when I do.

Thirdly: Max CPU Boost Clock Override which I usually set at like 125mhz or there about

Fourthly: Platform Thermal Throttle Limit: I use 89c

Lastly Curve Optimizer: Now this one is a little tricky, most places I read that you use different off-sets for different cores with the best cores usually lower (or negative higher) than you worse cores. Personally I use a flat figure for all cores. I figured I’ll first see what works and then I’ll work around with all the cores individually but in all honesty I’m not sure I’ll get to that soon. Now you want to select negative and first try like 10 or 15 and check that your system run and can bench. But use Hwinfo to monitor your core temps. I refuse to use AMD Ryzen Master as it is useless in that the figures it show is basically averages.

Now this optimized curve may seem stable and working for you but it may not be stable. I was able to do all the way up to the max negative 30 on all the cores but when I ran OCCT stress test it failed pretty quickly with errors. I ended up testing all the way from negative 10 all cores to negative 25 which all passed. Even negative 26 through errors so I am back to negative 25. Remember to test it to make sure you do have a stable system, OCCT stress test will run for about an hour.

Now a place where a lot of performance in left under the table is your FCLK or Infinity Fabric clock speed. This should be half the classified memory speed. As in 3200 will be 1600mhz and 3600 memory will be 1800 and if you running 3800 memory it must be 1900mhz. I’ll add a picture to show my timings with Zentimings as it would show all the details and what the FCLK and UCLK is.

Zentimings.jpg
 
Thanks to this, my CPU score went up by a huge amount!

CINE R23 - from 25k to 29k
3D Mark - from 14200 to 16200

So I do thank you so much! Temps are still a bit higher, by about 5 or so degrees, but will play around to see if I can lower the temps a bit...
 
Thanks bud, will try some of this later.

I am on the Asus B550 F Gaming and has some issues whereby my CPU wouldn't boost at all until I enabled the Asus Boost / Performance option. That seems to ignore a few of the other settings I had set.

Current loadout:
Temps +- 75C
All Core +- 4550
Single Core +- 5050
CB20 just under 11K
 
Thanks bud, will try some of this later.

I am on the Asus B550 F Gaming and has some issues whereby my CPU wouldn't boost at all until I enabled the Asus Boost / Performance option. That seems to ignore a few of the other settings I had set.

Current loadout:
Temps +- 75C
All Core +- 4550
Single Core +- 5050
CB20 just under 11K
Those temps, under load temps? My idle temps are around 45C and while doing CineBench runs, went up to 84C
 
Those temps, under load temps? My idle temps are around 45C and while doing CineBench runs, went up to 84C
Yes. I will run some testing again and post those, just to make sure.

After that I will apply the new settings and post results + actual settings changed.
 
Now a place where a lot of performance in left under the table is your FCLK or Infinity Fabric clock speed. This should be half the classified memory speed. As in 3200 will be 1600mhz and 3600 memory will be 1800 and if you running 3800 memory it must be 1900mhz. I’ll add a picture to show my timings with Zentimings as it would show all the details and what the FCLK and UCLK is.

I would also add here that you should enable PBO and keep XMP disabled for stability testing and only enable XMP and set FCLK once PBO is configured and stability testing is complete.
 
I would also add here that you should enable PBO and keep XMP disabled for stability testing and only enable XMP and set FCLK once PBO is configured and stability testing is complete.
This 👆 , I did not do LOL
 
I would also add here that you should enable PBO and keep XMP disabled for stability testing and only enable XMP and set FCLK once PBO is configured and stability testing is complete.
I prefer to have the memory at the speed I want it to run when testing stability as it forms part of the setup. I do not use XMP I set the memory speed where I want it with the timings I have tested and if it can pass Karhu memtest at 12000% or more I am happy. The OCCT test is for both memory and cpu but more so the cpu. It really show very quickly if your curve optimizer is not right.
 
I prefer to have the memory at the speed I want it to run when testing stability as it forms part of the setup. I do not use XMP I set the memory speed where I want it with the timings I have tested and if it can pass Karhu memtest at 12000% or more I am happy. The OCCT test is for both memory and cpu but more so the cpu. It really show very quickly if your curve optimizer is not right.
Currently my memory speed is set to 1900, but my XMP/DOCP or whatever it is called, is also active, should it be deactivated or not?
 
I prefer to have the memory at the speed I want it to run when testing stability as it forms part of the setup. I do not use XMP I set the memory speed where I want it with the timings I have tested and if it can pass Karhu memtest at 12000% or more I am happy. The OCCT test is for both memory and cpu but more so the cpu. It really show very quickly if your curve optimizer is not right.

XMP is vastly improved in DDR4, the problem was with DDR3 that the memory controllers on the older Intel chips were vastly reliant on the Silicon lottery and also power delivery. 4 DIMMs could make a big difference compared to 2 DIMMs if the higher XMP timings could work or not.
 
Currently my memory speed is set to 1900, but my XMP/DOCP or whatever it is called, is also active, should it be deactivated or not?
Don't worry about it, if you've set you memory and timings all good.
 
Excellent post.

Thank you @VPII

I will be following this thread
Thank you, In all honesty when I helped @HNO3 last night I realised there is a lot to share for those still messing around with AMD Ryzen 5000. In all honesty, I am no expert, I just shared what I found to work for me. That setting of PBO limits set to disable really works as it would give a drop of between 15 to 20c core temp without losing single core performance.
 
Just my additional two cents... My chip is able to perform without a performance hit at a negative offset of 0.1v, and helped me to stay sub 80c.. anything more than that and I start to lose clock speeds but it’s not much- there might be value in lower temps but also less power.. also people with Asus boards (not 100% about the TUF series or other brands) make sure that your ‘PBO Fmax enhancer’ is disabled as for me personally, this killed my effective clock speed for all core but did not affect my single core speed :)
 
Coincidentally I was messing around with this last night (and again now after seeing the post)

If I set the curve optimizer (all core) to any value then the machine either wont boot or at best hang at the login screen. I started at 30, 15, 10, 5 & 1... no joy
 
Thanks alot for this thread, still learning with regards to OCing CPUs and ryzen in general
I have applied everything you stated here but my score in 3d mark has barely increased


Am i missing any settings that could be keeping it in this range?

Does this have to do with me running my ram at 3200?
 
is this with AGESA 1.1.9.0 or 1.1.0.0 patch D?

cause not all vendors have 1.1.9.0 available

Edit:
i see its 1.2.0.0 on your ACE mainboard. Curious to see what the differences is with different AGESA
 
Last edited:
I prefer to have the memory at the speed I want it to run when testing stability as it forms part of the setup. I do not use XMP I set the memory speed where I want it with the timings I have tested and if it can pass Karhu memtest at 12000% or more I am happy. The OCCT test is for both memory and cpu but more so the cpu. It really show very quickly if your curve optimizer is not right.

I took your advice and setting XMP does hinder performance.

Check the clock frequency I scored 16 949 in Time Spy
 
Those temps, under load temps? My idle temps are around 45C and while doing CineBench runs, went up to 84C
Yep, checked the temps again now. Idle is about 31 and peak at 79.3 for CCD 1.

All core on CB20 is 4550, 9 of the cores hit 5050 in normal usage.
 
Seems like the 4550 was not a real reflection of the effective clock. It was more like 3900-ish.

Applied some new settings. Temps (idle/max): CCD (31/75) & CPU (36/75). Clocks are now 4360 All core & still getting the 5050 boost on single.

5950X with 64GB Vengeance Pro RGB 3200MT (4*16) on an Asus B550 F-Gaming. FULL list of settings (from default):
  • Asus Performance Enchancement --> Enabled
  • Memory frequency --> DDR4 3200MHz
  • DRAM CAS# Latency --> 16
  • Trcdrd --> 18
  • Trcdwr --> 18
  • DRAM RAS# PRE Time --> 18
  • DRAM RAS# ACT Time --> 36
  • TRC --> 54
  • VDDCR CPU Voltave --> Offset Mode
  • VDDCR CPU Offset Sign --> [-]
  • VDDCR CPU Offset Voltage --> 0.050
  • DRAM Voltage --> 1.37
  • Hot Plug --> Enabled
  • System in sleep, hibernate or soft off states --> Aura Off
  • USD Power Soft Off USB state S5 --> Disbaled
  • CPU Fan Speed Lower Limit --> 600rpm
  • CPU Fan Profile --> Turbo
  • AIO PUMP Control --> PWM Mode
  • Armoury Crate App --> Disabled
  • Precision Boost Overdrive --> Advanced
  • PBO Limits --> Disable
  • Curve Optimizer --> All Cores
  • Curve Optimizer Sign --> Negative
  • Curve Optimizer Magnitude --> 12
  • Max CPU Boost Clock Override --> 200MHz

iCue settings on AIO pump: Extreme (fans are custom curves)
 
I took your advice and setting XMP does hinder performance.

Check the clock frequency I scored 16 949 in Time Spy


Had to disable PBO and FCLK today, PC keeps rebooting randomly.

Hit the CMOS reset button and everything is stable.

On power save mode currently and I need to find out what was causing this. Was stable for a few days up until today.
 
Had to disable PBO and FCLK today, PC keeps rebooting randomly.

Hit the CMOS reset button and everything is stable.

On power save mode currently and I need to find out what was causing this. Was stable for a few days up until today.
In all honesty, the curve optimizer setting may seem stable but could very well not be stable. The best way to test is to set it and run OCCT stress test which takes 1 hour. Doing so if it pass would mean your setting is stable or there about. I've fallen back to -10 all core and ended up testing till -25 all core which passed so that would be it as -26 was a no go with errors.
 
Anyone here tried Clocktuner for ryzen and do a comparison betwen all these manually? because CTR and Dram calculator did most of this and good gains compared to me doing everything manually in bios

I do have the Ryzen 3600 though but CTR does support 5000 series,

would be interesting to sees the difference on 5000 series
 
Just to add a result I got with my system running PBO and -25 core off-set. I had some issues yesterday that my system won't even startup. What happened was I installed MSI Live to get the latest bios for my MSI Gaming X Trio GPU but it also downloaded some other software and drivers which it felt I needed. One of which was MSI fast boot and upon installing it it prompted me that it needs to restart to finalise the setup and upon restarting it won't restart hanging Qpost 4d. I tried many things to get it to work but nothing helped so I left it off for a while. When I started it up again it magically worked. Reflashed the bios just in case and all good now. So I decided to just check performance, could not get my memory working properly with latest bios so I fell back to the one before. But I was pleasantly surprised with the result I got in CB20

CB20-PBO-Best-3800mem.jpg


CINEBENCH-R23-CPU-Multi-Core-30355.jpg
 
So I had a bit of an issue over the past couple of days as my PBO clocks were sitting around 550mhz, yes 0.55ghz. I tried everything but nothing seems to have worked, from flashing back to the first bios released for AMD Zen 3 cpus which actually worked but my memory clocks would not go more than 3200 which is the XMP speed.

I contacted MSI explaining to them my situation, after flashing back to the latest bios for the MSI Meg X570 Ace motherboard. The got back to me and stated that I need to press F6 in the bios to load optimized defaults and what do you know. Upon doing this I am back at almost normal PBO clocks of around 4.5 to 4.6ghz all core but fluctuating a bit so performance not the same as when running 4.5 ot 4.6ghz manual overclock.

The problem I had with their latest bios is that there is no option in the normal screen to do a manual overclock. You need to search and under advance you get the overclock option where you state the speed and vcore, but the vcore you state there is not the one that actually apply. In the main screen you still need to set vcore to override and then set the vcore you want. Here is where I found that I can run 4.55ghz at 1.275vcore without an issue. No not prime stable, but it works for everything else. I settled for 4.625ghz using 1.3vcore as load temps running Cinebench R23 is around 74c which I am happy with.

System all good now and I might actually leave my manual OC running as temps compared to PBO is around 5 to 8c lower and idle temps between 25c and 30c.
 
I contacted MSI explaining to them my situation, after flashing back to the latest bios for the MSI Meg X570 Ace motherboard. The got back to me and stated that I need to press F6 in the bios to load optimized defaults and what do you know. Upon doing this I am back at almost normal PBO clocks of around 4.5 to 4.6ghz all core but fluctuating a bit so performance not the same as when running 4.5 ot 4.6ghz manual overclock.

Thank you for the info, I have updated to latest and reset to optimized defaults.

Going to try PBO and XMP again and go from there.
 
Thank you for the info, I have updated to latest and reset to optimized defaults.

Going to try PBO and XMP again and go from there.
Shot, let me know what happens. Normal PBO with a 5950X would mean around 4400mhz all core boost, when using the curve optimizer you can get up to 4.5 to 4.6ghz but fluctuating a lot.
 
Shot, let me know what happens. Normal PBO with a 5950X would mean around 4400mhz all core boost, when using the curve optimizer you can get up to 4.5 to 4.6ghz but fluctuating a lot.

I have just enabled PBO and the curve optimiser -30 for undervolting.

Leaving XMP disabled and everything else stock.
 
I have just enabled PBO and the curve optimiser -30 for undervolting.

Leaving XMP disabled and everything else stock.

5900x stable with PBO and XMP enabled with curve optimiser set to -25.

Getting all core boost to 4375mhz.
 
5900x stable with PBO and XMP enabled with curve optimiser set to -25.

Getting all core boost to 4375mhz.
If you want to know if stable run the latest OCCT the one hour test, it will show you if errors then you know the curve optimizer -25 need to be lower, or higher as in -20 or maybe even -15.
 

Users who are viewing this thread

Back
Top Bottom