Overclock.net banner
Cancel
Create thread New Forums

PBO Enabled + Auto OC OR PBO Disabled: And Every Other OC Tool in the Bios

Jump to Latest
1 - 20 of 38 Posts

EastCoast

· Banned
Joined
·
5,701 Posts
Discussion starter · #1 · (Edited)



This is what PB/PBO Controls:


EDC
After watch Gamer's Nexus review in how they test these cpus I now know it's either one or the other in the title of this thread.
Therefore, PBO and or PB is starting to look pretty pointless endeavor as it only allows the CPU to if, and only if, certain thermal thresholds/power requirements are met.



As a whole these cpus overclock/boost like GPUs. If you are already getting the max boost of the CPU, tweaking PBO doesn't let you go past that (based on the features listed above alone).




GN: Changing these values to avoid reaching the limits of PBO. It is not suggested that you copy this.

You can find out what your MB sets those values to by enabling PBO in the bios. Get to desktop and start Ryzen Master.

PPT: XXX
TDC: XXX
EDC: XXX
How the above values are set are based on the motherboard used. As 2 different motherboards will provide 2 different sets of values. Higher values have the potential of disabling these features as CPU/MB would never reach those limits. Lower values are used to reach AMD Spec.





The offset used hoping that you can overclock to 200+ offset is actually Auto Overclocking and it is not part of PBO even though it's found in the same sub menu. But used in conjunction with PBO to help the CPU OC better during heavy threaded workloads. The problem is that this doesn't work as good as you think it should. Furthermore, doing this doesn't guarantee 200Mhz offset when you set the offset that high. As the base frequency you are basing this off of is variable.


Test Results







PBO Max = PBO using motherboard limits in the OC section > PBO of the bios.



The above spoiler shows you there is no difference between PBO off vs PBO on vs 200Mhz offset. Which is another name for Auto OC.





Furthermore, it seems that PBO seems to only scale highest performance when you are using liquid nitrogen (around -56C provided roughly 200 point bump in cinebench15 from an average of 80C). As Steve points out towards the end of the video starting around 29:21 mark of the video. As every couple degrees of change can have an impact on actual performance with PBO. Ryzen CPUs are sensitive to ambient and overall temps.




31:03 Mark of the video
The limits that PBO raise aren't the limits encountered in testing/gaming (unless you are using liquid nitrogen). The limit most important is temperature. Perhaps, IMO, temps. are too important. In order to get that 200Mhz offset you need an average temp on the CPU of 55C with PBO disabled!!!!!
This is not including the curve. Just PBO.



 
#2 ·
The 200MHz offset adjusts the "Frequency Limit - Global" value in HWiNFO. I've done my own testing to validate this. As my 3800X is locked by default to 4,550 MHz. However, if I go into the PBO settings and include the 200 MHz offset, i'll see that value go up to 4,750 MHz. It doesn't mean my actual reported boost will hit it. And from my experience, it never does. I've only ever hit 4,600 MHz.

2487772

And the "Idle Downclocking" is a bit misleading with that chart. As manual overclocking still enables downclocking /idling.

See screenshot below with my manual OC at 4.35GHZ.... Look at temps and power. In all my testing, its the exact same behaviour as Auto settings or PB/O settings.

2487773

Edit: I see you're updating the thread / OP but hope this helps some users.
 
Discussion starter · #3 · (Edited)
Slaughtahouse said:
The 200MHz offset adjusts the "Frequency Limit - Global" value in HWiNFO. I've done my own testing to validate this. As my 3800X is locked by default to 4,550 MHz. However, if I go into the PBO settings and include the 200 MHz offset, i'll see that value go up to 4,750 MHz. It doesn't mean my actual reported boost will hit it. And from my experience, it never does. I've only ever hit 4,600 MHz.

View attachment 2487772

And the "Idle Downclocking" is a bit misleading with that chart. As manual overclocking still enables downclocking /idling.

See screenshot below with my manual OC at 4.35GHZ.... Look at temps and power. In all my testing, its the exact same behaviour as Auto settings or PB/O settings.

View attachment 2487773

Edit: I see you're updating the thread / OP but hope this helps some users.
Click to expand...
Yes, you are right you do get something if you aren't already reaching max. I relabeled my title to include the fact that you are already reaching max boost without it. Would using PBO make a difference?

Edit:
However, what appears to be a major factor in how to get the best performance from Ryzen CPUs is how cool (within power limits) you can get your cpu. Regardless of what CPU frequency you obtain. Which is why you see some using a curve to undervolt the cpu.

Another interesting observation is that I've seen apps like MSI AB, HWInfo, etc tell me I'm at 4.850GHz. However, Ryzen Master is telling me I'm around 4.4GHz while gaming. So...there's that.
 
#4 ·
EastCoast said:
Yes, you are right you do get something if you aren't already reaching max. I relabeled my title to include the fact that you are already reaching max boost without it. Would using PBO make a difference?

Edit:
However, what appears to be a major factor in how to get the best performance from Ryzen CPUs is how cool (within power limits) you can get your cpu. Regardless of what CPU frequency you obtain. Which is why you see some using a curve to undervolt the cpu.

Another interesting observation is that I've seen apps like MSI AB, HWInfo, etc tell me I'm at 4.850GHz. However, Ryzen Master is telling me I'm around 4.4GHz while gaming. So...there's that.
Click to expand...
The best clocks I got were...

Single Core: 4,600MHz (PB/O) vs 4,550MHz (stock);

All Core: 4,300 to 4,400 MHz on a single CCX / 4,100 to 4,200 MHz on the other CCX (PB/O) vs. 4,150 MHz all core (stock).

However, temperatures were inconsistent and hitting 75C at various times. I don’t recall the TDC / EDC settings for these “best” results.

Scalar I’ve tried 1x, 2x, and 10x.
I’ve tried motherboard limits.
I’ve tried various TDC / EDC settings
I’ve tried the “EDC Bug” (EDC set to 1)
I’ve always kept 200 MHz offset.
I’ve also tried various voltage offsets including: approx -.050, -.1, and -.15v

I never really understood what I was doing wrong or how it works. That’s why I opened up your thread.

I watched that video and I’ve also read this:
www.techpowerup.com

Zen 2 OCing... The right way?

No TL/DR, read or don't read, w/e not bothered either way. Overclocking Zen 2 is not as straight-forward as some would like to believe, it is not an Intel, so stop treating it like one. I am not a fanboi of either camp, so please don't think I'm bashing Intel methodology, it's simply...
www.techpowerup.com www.techpowerup.com

AMD Ryzen 3000 Undervolting Offset vs. Override | Vcore Voltage | GamersNexus

stub This is a quick and straightforward piece inspired by a Reddit post from about a week ago. The reddit post was itself a response to a video where a YouTuber claimed to be lowering temperatures and boosting performance on Ryzen 3000 CPUs by lowering the vcore value in BIOS; we never did...
www.gamersnexus.net www.gamersnexus.net

www.overclock.net

Why you should undervolt your Ryzen 3000 - Part II

Hear me out! I know we had this theory before! Thanks to The Silt writeup on FIT and VFT curves, i finally understand AMD have made Ryzen 3000 closer to GPU boosting than old Intel Turbos. The biggest problem with initial testings, the reviewers did not fully unclock the PBO motherboard limits...
www.overclock.net www.overclock.net

In my very best attempt, I followed the PBO + BCLK method and had my bus running at 102MHz, resulting in a single core ~4,650 MHz boost.

I loaded up Rainbow Six: Siege and ran the benchmark 3 times and results were identical to my all core, 4,275 MHz / 1.176v all core OC.

It’s also a similar conclusion to most threads I’ve read. At the end of the day performance improvement is negligible.

So because of all of my experience, I just play with all core OC.

I don’t have to worry if my CPU will boost.
I have full control of the voltage (limit).
My power consumption under load is lower.
My CPU heat is lower in AIDA 64 / CPU benchmarks. Approx 14c lower...

I’m still open in learning how to figure out PB/O but I personally can’t recommend considering the performance, heat, and power consumption vs. the effort it takes.

The most versatile tool I have used for overclocking Zen 2 is CTR 2.0 (by far).

In Hybrid OC mode, I could have the following settings:
P0 = Stock single core boost @4,550 MHz (stock voltage)
P1 = Light / Gaming load @ 4,350 MHz 1.25v
P2 = All core load @ 4,275 MHz 1.176v

The ONLY thing I don’t like about CTR is that it is another application running. If you’re willing to let it run, it offers in my mind, the most straightforward approach to getting the behaviour you want. As well, the other downside is you can’t modify single core behaviour (at least with the version I have on Zen 2).

Hope this helps provide some context to my experience.

Note: In no way does the above mean this is the only method of OCing Zen 2. Everyone has different goals, preferences, and most importantly... hardware.
 
Discussion starter · #5 ·
Thanks for the insight.

So apparently Gamer's Nexus simply
-Opened Advanced AMD OC'ing
-Set Percision Boost Overdrive to Advance
-Set PBO Limit to Motherboard
-Set Max CPU Boost Clock Override to 200Mhz
-Tested
 
#6 · (Edited)
This seems out of date now, Zen3 PBO is very different from Zen2 PBO in my experience.

Slaughtahouse said:
So because of all of my experience, I just play with all core OC.
Click to expand...
On Zen2, I agree. With Zen3 though PBO works much better, I can get a very high all-core while still seeing 5050 MHz sometimes in single core. Trying for an all-core OC will limit single core performance a lot while improving all-core performance very little.
2487801


This is using the motherboard power limits, hitting the 200A EDC limit, not opened up completely. Temps are above 80°C during multicore so there isn't much room for more power given my current cooling.
 
Save
Reply Quote Rep+ Rep+
Discussion starter · #7 · (Edited)
@Slaughtahouse
I've been reading a whole lot about this and been testing this myself. And you are right. The best option is an All Core OC.
However, I ran into a snag doing that. After several runs of Cinebench 15 and 20 at 4.850GHz @ 1.30V my scores are lower (about 67C) then at 1.425V (78C-81C depending on room temps). Which is the default voltage...although it rarely reaches that high in games which is about 1.32V. Which is the same thing Steve at GN posted a video on regarding something about only using 1.00v. Meaning that you might get away with using 1v but actual performance will suffer.

Which can induce clock stretching. The funny thing is that most monitoring software gives the impression everything is fine when it's not as your performance results will degrade.
IE: At 1.0v you can boot into win10, test games and have MSI AB show you are at 5GHz. But the actual performance is much lower. This is why I am now using Ryzen Master (enable clock monitor, etc) to monitor actual CPU frequency as it appears to be accurate to what the actual performance is showing.

Therefore, the best way to OC is the ability to do an All Core OC that can go as high as 1.425V in game but also be able to go into a Power Balance mode that can reduce power consumption and clocks at 2D mode. And, the only way I know that can happen is if we petition AMD to allow it in a revised Agesa Update.

This is the only way I can see getting the best from both:
-Single threaded results
-Multi threaded results
Which can provide an actual performance boost. While keeping temps under control without worrying about runaway thermals at 2D.
IMO

In other words I'm not pressed to show you I'm getting 5GHz All Core. But actually show you I'm getting a boost in performance that is equilivant to 5GHz all core.

Edit:
But you know what really confused me the most?
I found Precision Boost Overdrive in 2 separate areas of the bios with 2 set of features.

Under Extreme Tweaker Tab there is PBO that offers:
PBO Fmax Enhancer
Precision Boost Overdrive
Precision Boost Overdrive Scalar
Max CPU Bost Clock Override
Platform Thermal Throttle Limit

Under Advance > AMD Overclocking there is a PBO Option that offers:
Precision Boost Overdrive
PBO Limits
Precision Boost Overdrive Scalar
Curve Optimizer
Max CPU Boost Clock Override
Platform Thermal Throttle Limit

What the heck man?? What the heck????
😕😳:oops:
 
#8 · (Edited)
I don't use auto OC. Actually I dont use ryzen master at all. Trying to keep as little programs on my pc as possible. But I do dial in PBO + manually set my max core voltage @1.28. It produces lower cinebench scores, but in lighter loads like gaming, I see much higher boost frequency and duration.... more stable and higher frame rate. As much as 5.1ghz but typically between 5 and 5.05 ghz on 5800x. I do set PBO limit to motherboard. +200 override, and somewhere around -26 on the curve. Depending on which core likes what. Temps stay around 52C in games and I'm using cheap GD900 thermal paste (soon to be upgraded)


Originally I did manually OC all cores to 4.8 permanenetly, then I used auto OC which yielded better in game results than locking to 4.8, but the PBO alone is noticeably better.

Yeah, the Gigabyte bios leaves some to be desired, lol. I always go thru AMD overclocking

On another note, I've found that the provided AMD chipset driver from Gigabyte is somehow worse than what windows automatically installs.
 
Discussion starter · #9 ·
Insidious Supra said:
I don't use auto OC. Actually I dont use ryzen master at all. Trying to keep at little programs on my pc as possible. But I do dial in PBO + manually set my max core voltage @1.28. It produces lower cinebench scores, but in lighter loads like gaming, I see much higher boost frequency and duration.... more stable and higher frame rate. As much as 5.1ghz but typically between 5 and 5.05 ghz on 5800x. I do set PBO limit to motherboard. +200 override, and somewhere around -26 on the curve. Depending on which core likes what. Temps stay around 52C in games and I'm using cheap GD900 thermal paste


Originally I did manually OC all cores to 4.8 permanenetly, then I used auto OC which yielded better in game results than locking to 4.8, but the PBO alone is noticeably better.

Yeah, the GB bios leaves some to be desired, lol. I always go thru AMD overclocking
Click to expand...
Hmm, thanks for posting this. Do you have an auto OC feature in the bios? I haven't found an Auto OC feature in the bios.

So Games can be consider light loads. While Cinebench is considered Heavy Loads. It's by no means an absolute but does explain how the CPU is working.

While watching those GN video Steve considers Auto OC using:
-Max CPU Boost Clock Override: 200Mhz
-PBO Limits: Motherboard
I'm not sure what else is involved though. I've only seen Auto OC in Ryzen Master which 'appears to mess with the bios' requiring a reboot in order to activate Auto OC.
 
#10 ·
Hm, ok, good to have the clarification.... I was actually using auto OC in Ryzen master at one point... Actually because it worked better than PBO on an older bios version.... It seems to sort of enable PBO? maybe orerriding bios.... I'm not sure.... But now I'm curious if auto OC works better with the new bios....

What I noticed with auto OC is that boost duration was very poor. Night and day versus dialing in PBO it's-self and an under-volt if at all possible because it will use thermal headroom to dial up boost speed and duration.
 
Save
Reply Quote Rep+ Rep+
#11 ·
Thanks for the context Asmodian and East Coast

Also, regarding clock speed in software. I read and read and re-read that HWiNFO post from the author. You have to be vary careful where you’re reading your clock speed.

Most software, to my knowledge like MSI Afterburner will tell you clock speed by taking the Core Ratio x Bus Clock.

However, for Zen and even more recent Intel CPU, that way of reporting is no longer accurate. I swear only by “effective clocks” and I urge others to read up.

In short, the effective clock reading is polling info over a set duration and averaging it.

So if your cpu core heads into various c-states, sleep states, is parked etc, that info is averaged over the poll rate. E.g. If your CPU boosts from 1000 MHz to 200 MHz to 0 MHz to 0 MHz within one second, effective clock will capture than and average it. So it will show 1200 MHz / 4 = 300 MHz.

FEATURE - Effective clock vs instant (discrete) clock

It has become a common practice for several years to report instant (discrete) clock values for CPUs. This method is based on knowledge of the actual bus clock (BCLK) and sampling of core ratios at specific time points. The resulting clock is then a simple result of ratio * BCLK. Such approach...
www.hwinfo.com www.hwinfo.com

I assume both Ryzen Master and HWiNFO are the only software to capture this data on Zen.

I still see MANY people on this forum post their speeds from the “core clocks” in HWiNFO. Including myself, I didn’t understand it until I found that post from the author earlier this year.
 
#12 ·
Cinebench is a bad way to test PBO, you are hardware/software limited on your AVX instructions for allowed EDC. It will throttle. (why less voltage boosts better)
The EDC setting in BIOS is only for "non-AVX" loads. Only the PBO bug allows you to bypass this limit on Ryzen 3000 processors.

Test other instructions sets for load when you test PBO if it allows you to boost higher with all tweaks enabled. You will see a significant better effect than running Cinebench which is still locked down with a fused limit of allowed EDC to be used.
Basically Cinebench is a load that will always throttle because it hits the limit on EDC it can run for the processor sku:s avaible with PBO.

The 3900/3950X are significantly gimped because of this fused limit.
The 8 core 3800X has most headroom while the 65W 3700/3600 have lower limits.
 
Save
Reply Quote Rep+ Rep+
Discussion starter · #13 · (Edited)
It appears that the actual clock rates of these new ryzen CPUs is between 4.0 and 4.4 GHz all core using ryzen master. However, other monitoring apps show 4.850 or 5 gigahertz that you set it to. Which isn't right as slaghtahouse suggests.

I can only ask that those thinking they're getting frequencies that high with ryzen 5000 to use ryzen master's cpu monitoring feature and see what it reports while gaming.

If there's anyone who reads this post that is actually getting 4.850GHz or better could you either post a video or a animated gif of ryzen Master cpu monitoring showing those all core frequencies. And, could you please explain what you did to get there.

We can tweak the BIOS to reflect 4.850 gigahertz or 5 gigahertz. What is actually happening while gaming appears to show clock rates far lower when using ryzen Master. Which I have found out to be the most accurate representation of the clock rates that you are getting
 
#14 ·
Are you expecting someone to get 4.85 GHz or 5.0 GHz all-core? Because that is crazy on Zen3. Does anyone say they can get 5 GHz all core?

I see 4.950+ GHz in Ryzen master for single core in real workloads, but not multi-core. Multi-core seems to top out at about 4.750 GHz with a lower power workloads using my PBO settings, dropping to 4.550 GHz with higher power loads (200A).

The reason to use PBO is not to get the highest all-core OC, it is to get a good all-core OC while also keeping at least stock single core performance (Zen3 pushes single core very hard at stock, at least on my 5950X).
 
Save
Reply Quote Rep+ Rep+
Discussion starter · #15 · (Edited)
Update:
I've completely disabled PBO from both Extreme Tweaker and Advance. Set PLL to 3. Used Ryzen Master to enable Auto OC:
Hmm, enabling Auto OC in Ryzen Master changes the bios settings for Advance >AMD OC'ing> PBO :
PBO: Advance
Max CPU Boost Clock Overide : 100Mhz



With the new Radeon 21.4.1 drivers I can now do some cpu monitoring.
This is what I'm seeing in Valhalla @ 1440p. Game is very smooth though. But my CPU temps are very low. Not sure what to make of it as the timeline shows a steady 45C.
But I'm averaging around 4.4GHz to 4.5GHz

Using Advisor, in radeon as a cursory result, minimums appears to have improved. As I never got excellent in this game before.
In game I've mostly seen 95FPS+ based on the area I was in.
 
#16 ·
Asmodian said:
Are you expecting someone to get 4.85 GHz or 5.0 GHz all-core? Because that is crazy on Zen3. Does anyone say they can get 5 GHz all core?

I see 4.950+ GHz in Ryzen master for single core in real workloads, but not multi-core. Multi-core seems to top out at about 4.750 GHz with a lower power workloads using my PBO settings, dropping to 4.550 GHz with higher power loads (200A).

The reason to use PBO is not to get the highest all-core OC, it is to get a good all-core OC while also keeping at least stock single core performance (Zen3 pushes single core very hard at stock, at least on my 5950X).
Click to expand...
I can easily get mine to BOOST past 5 on all cores for a significant amount of time. I can also lock mine to 4.8 permanently as far as I've tested.
 
Save
Reply Quote Rep+ Rep+
#17 ·
Insidious Supra said:
I can easily get mine to BOOST past 5 on all cores for a significant amount of time. I can also lock mine to 4.8 permanently as far as I've tested.
Click to expand...
Wow! Core effective clocks? What do you have for cooling?

What voltage for 4.8? :)
 
Save
Reply Quote Rep+ Rep+
#18 ·
Yes, core clock locked to 4.8, doesnt fluctuate at all. I havnt tried it again since newest bios, but it seems I get better results using PBO and boost to 5.05-5.1. I think I'll test again here in a few to see if there is a performance gain locking or not. I'm just measuing performance in games (1440 ultra) which is my typical workload for the CPU

I think I was runing 1.26V Currently 1.28 which sounds silly.... But theres also the factor that my Gigabyte board is likely under-reporting voltage, I think. Which wouldnt be surprising... shady business practice to attempt better performance figures... Which would also be crazy because my board defualts to something like 1.46V!!! And the system crashes like crazy like that. HWinfo says there is likely a %75 accuracy for the power readings it sees.

I have an Arctic 280mm AIO that I lapped to my IHS, polished both afterwards using both again to polish eachother, and only using GD900 thermal paste right now. Will switch to kryonaut here very shortly. (y)
 
Save
Reply Quote Rep+ Rep+
#19 · (Edited)
Ok, so I can't get my game to boot without a crash with all cores locked to 4.8 anymore. Tried a lot of different cpu voltages. Likely due to increased ram/fclk. I'm pretty sure I was running 3600/1800 prior, and on f13a bios. Now at 3800/1900. I can get as high as 3933 and 1966 as far as I've tested, but frame rate peaks and lows intensify although average goes up. Same with 1933 vs 1900.

update fooling with ryzen master bugged my bios out somehow although I reverted back to same settings. Luckily I know all the right inputs for stable settings by memory now, lol. Had to reset bios. Admitedly... gigabyte bios is a bit iffy if you ask me.... it has duplicate settings for several things like PBO and clock speeds, ram settings, in different areas.... I set my fclk in a different area this time, and my screen wouldnt even turn on lol. Had to reset again and set fclk in AMD overclocking section.
 
Save
Reply Quote Rep+ Rep+
1 - 20 of 38 Posts
About this Discussion
37 Replies
6 Participants
Last post:
EastCoast
Overclock.net
posts
28M
members
562K
Since
2004
The premier forum for overclocking experts and enthusiasts. Discuss hardware optimization, custom builds, benchmarking, cooling solutions, and pushing the boundaries of computing performance. From beginner guides to extreme overclocking, join our technical community to master system tuning.
Full Forum Listing
Explore Our Forums
Intel CPUs Hardware News NVIDIA PC Gaming Water Cooling