AMD Curve Optimizer any guides / experience

[Zach_01]

1. You should run a benchmark or check the effective clock frequencies to compare. Just a theory, but it's possible you will get a slight degradation in performance if the cpu tries to boost higher, uses more power, and then lowers the effective clocks as a result. If that happens it is likely only a small effect though.


2. This is something I have been wondering about. There was another post somewhere saying this was actually enabled in Zen3, but I don't know if that was just speculation or not. Assuming there is no documentation or other statement from AMD that they have enabled individual core voltages on desktop then I'll continue to assume all cores run off a single voltage.

3. One thing I realized, if the cpu has voltage regulators on die, then are the seemingly high (1.4-1.5) voltages what is actually seen by the cores? It would explain a lot to me if the cpu voltage we are seeing is actually the supply to onboard voltage regulators and not the actual voltage that the cores run at.


4. There's multiple things at play here, but the short answer is that it is not pointless to optimize on a per-core bases. At least if you want to get the last few % and as far as my understanding goes:

_1. -5 on your best core might actually result in the same VID as -30 on your worst core, since each core has a different voltage curve set from the factory (and curve offset is setting an offset to the factory default).
_2. Core VID in HWInfo is the voltage requested by that core. The actual voltage is going to be different to that, and assuming there is only a single voltage set for all cores it will be based on the highest voltage requested among the cores and possibly other parts of the die.
_3. For a multi-core load the voltage will be set based on the highest voltage requested among the active cores. You can focus on those cores when tuning curve offset to save time, but you need to figure out which ones they really are first.


5. Also note that by default HWInfo will report the same VID for all cores, but will read out sensor/register values sequentially, so in your screenshot you might just be seeing the VID changing between the time it read the value for Core 0 and Core 1. There is a setting that was added in HWiNFO v6.40 which seems to fix this. If you enable this I'd expect that you will see a slightly different VID for each core:

• Added Snapshot Polling mode for AMD Zen-based CPUs.

1. Yes the effective clock shows the sustainability of the boost/clock. In other words, the true(er) speed of a core.

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

2. (post #13)

Normal vs. Snapshot polling mode

Hi Martin, and everybody here! Here is screenshot of my PC running Cinebench R23 MT test. I've merged two sensor panel screenshots from different modes. Can someone explain, why reported core VID's values are - all the same and close to Vcore SVI2 value (normal mode - left panel) - all...

www.hwinfo.com

3. The voltages of 1.4~1.5V is what the board VRM are supplying. At least on my R5 3600/AorusPro X570 system I can verify this via the VR sensor readings (VR VOUT).
Anyone with VR VOUT readings and 5000series CPU? ...shed some light please, although I expect it to be the same as 3000series (see 2.)

4. I too don't think it's pointless. The individual offset could just mean different end speed for each core at the same voltage, but that is not something I can verify as I don't own a 5000 CPU. And this is making sense if you think about how the core speeds are acting under stock conditions. Why the curve optimizer would work any different? It is just a voltage/speed curve alteration over stock.

5. As matter of fact, after I enabled it I see identical VIDs (cur/min/max/avg). Prior to this there was a difference in requests. And again, I own a 3000series CPU.

---------------------------

Lets not forget that there is no software that can monitor and report speed/voltage 100% accurately when those values are different every 1~20ms (depending power plan), and a software's polling period is 500, 1000, 2000ms.
By the way, dont use polling period under 1000ms on HWiNFO because it will probably keep the CPU more on active state (and on higher voltage) and out of halt/sleep states.

 

[blu3dragon]

At this point I don't really know what I'm doing anymore, there are just too many questions and I'm totally sick of prime95 by now

Should I test the individual curve offsets with all other curves at default or is it perfectly fine to do so with some offsets already set?

Does the increased frequency limit even matter when (according to someone on a different forum) it adds "up to" 200MHz, meaning it's more like +100MHz with two cores and probably next to nothing with more than that running under load? Just by having programs opened and browsing the internet, I see activity on four cores most of the time.
and
Is the frequency limit of any use when I'm not overclocking in the classical way of increasing power limits, voltages etc.?

Can/does the increased frequency limit affect how much of an offset can I use?


Re: threads. I am not sure. I've always - for years - read that for HT/SMT enabled CPU, you have to run two threads, otherwise the load won't be 100%. I don't want to dig into reading about any of the technical klingon behind that, so I just do what I've been doing for years.

OK, the first thing I want to say is that this is all for fun. If you are tired of testing or not having fun anymore, then take a break In the overall scheme of performance of your PC we are really only chasing small gains here and in the end it's probably not worth doing other than for the fun of it as AMD has already tuned the CPU to work very close to its limits out of the box. If you want the simple solution, just enable XMP, enable PBO (but keep all PBO settings at default), run a few stability tests and check temps, and then just enjoy using your PC.

With regards to your questions:
1. It's perfectly fine to test with different offsets set on each core. Actually this is how I think you need to do it since you want to end up with some offset set for each core. Take a look at the "Example Scenario" in this post for some idea of how to go about this: https://www.overclock.net/threads/s...script-for-zen-3-curve-offset-tuning.1777112/
2. The increased frequency limit might not make much or any difference in any real load. For a 5600x it is probably worth trying for +200. For a 5800x it might not be worth increasing at all since,
3. Yes, if you increase the frequency limit it can affect how much curve offset you can apply, which will then reduce the maximum boost frequency for that core.

1. Yes the effective clock shows the sustainability of the boost/clock. In other words, the true(er) speed of a core.

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

2. (post #13)

Normal vs. Snapshot polling mode

Hi Martin, and everybody here! Here is screenshot of my PC running Cinebench R23 MT test. I've merged two sensor panel screenshots from different modes. Can someone explain, why reported core VID's values are - all the same and close to Vcore SVI2 value (normal mode - left panel) - all...

www.hwinfo.com

3. The voltages of 1.4~1.5V is what the board VRM are supplying. At least on my R5 3600/AorusPro X570 system I can verify this via the VR sensor readings (VR VOUT).
Anyone with VR VOUT readings and 5000series CPU? ...shed some light please, although I expect it to be the same as 3000series (see 2.)

4. I too don't think it's pointless. The individual offset could just mean different end speed for each core at the same voltage, but that is not something I can verify as I don't own a 5000 CPU. And this is making sense if you think about how the core speeds are acting under stock conditions. Why the curve optimizer would work any different? It is just a voltage/speed curve alteration over stock.

5. As matter of fact, after I enabled it I see identical VIDs (cur/min/max/avg). Prior to this there was a difference in requests. And again, I own a 3000series CPU.

---------------------------

Lets not forget that there is no software that can monitor and report speed/voltage 100% accurately when those values are different every 1~20ms (depending power plan), and a software's polling period is 500, 1000, 2000ms.
By the way, dont use polling period under 1000ms on HWiNFO because it will probably keep the CPU more on active state (and on higher voltage) and out of halt/sleep states.

Not sure if I have VR VOUT on my board? Here are a couple of screenshots that show different VIDs on a 5800x.

First one is p95, 16 threads, FFT size of 84 in place, avx2. Average voltages best captures the difference between cores.
Second one is captured while running the script, single thread, FFT size of 84 in place, avx disabled. Look at max voltages for this one.

 

[Det0x]

Today i found out how much latency difference there really is between a 1CCD and 2CCD Zen3 CPU
(5600x + 5800x VS 5900x + 5950x)

I disabled one CCD on my 5950x to simulate a 5800x
1 CCD = 51.7 ns in aida64
2 CCD = 54.2 ns in aida64

Running 4x8gigs bdie memory and PBO CO @ standard 24/7 everyday settings

 

[agentnathan009]

It is not how you do things. Also, 1.538v - I hope it is due to a fault in the sensor.
If you are looking for ST scores you clamp PPT, and otherwise for MT you keep PPT just 10% above EDC.
Those scores are too high. Better safe than sorry.

Ryzen 9 5950X Curve Optimizer to 5.1 GHz, PBO and overclocking - ElmorLabs

The much anticipated Zen 3 architecture has finally been launched close to a year and a half after Zen 2 while remaining on the same 7nm TSMC process. AMD's recently released Ryzen 3000XT showed there are frequency gains to be had just by manufacturing improvements over time. The question is...

www.elmorlabs.com

Your comprehension of how electricity works is sorely lacking. 1.5 volts is not an issue for CPU. Do high voltage transmission lines burn up carrying 200,000 volts? No, with an exception, when an object gets close enough that electrons jump from transmission line to grounded object, then electrons making that jump can cause a fireworks display and damage from electrons crossing the charged gap. Volts don't create heat, amps create heat. That is why transmission lines can handle 200,000 volts but are transmitting a few amps. Batteries, on the other hand, operate at low voltages, but with high amperage (current). When a battery is dead shorted the 100+ amps can cause the cable to get hot, start glowing red, and/or otherwise burn up. You CPU operates more like a battery than a high voltage transmission line.

The conclusion to this matter is that volts don't cause the CPU to fail, though I suppose that if they could get high enough and there was some less insulated part within the CPU that allowed the electrons to jump (see transmission line example) then volts could be responsible for killing CPU. Amperage is what makes your CPU hot. Find out what amperage your CPU can handle and work with that. Current CPUs have so many sensors and monitoring capabilities that they protect themselves, for the most part, from being burnt up due to overclocking.

 

[mtcn77]

Your comprehension of how electricity works is sorely lacking.

The conclusion to this matter is that volts don't cause the CPU to fail, though I suppose that if they could get high enough and there was some less insulated part within the CPU that allowed the electrons to jump (see transmission line example) then volts could be responsible for killing CPU. Amperage is what makes your CPU hot. Find out what amperage your CPU can handle and work with that. Current CPUs have so many sensors and monitoring capabilities that they protect themselves, for the most part, from being burnt up due to overclocking.

You seem awfully insecure for someone without even a trace 'notion' of how electricity travels in the opposite path that electrons travel...
The issue is, I 'know' how electrons are even shaped like and they can lose track into unobserved paths since they are not shaped like a 'single' orbital physical ring, but double orbital rings with a second perpendicular one connected at the electron's moment center that can crossect without any discrimination for physical barriers - its quantum orbit does not interact with physical matter, just cuts across it. So much to say, electrons having an extra invisible quantum orbital ring that they can always switch between just as easily as the one they are physically travelling on - that is the very definition of Murphy's Law in electricity "what will go wrong, does" without you even knowing it.
Your understanding of electrons is just half of the story, you are trying to solve it as a solid matter. That is not how it intuitively behaves.
But as the famous physicist Richard Feynman once said, “It doesn’t matter how beautiful your theory is, it doesn’t matter how smart you are. If it doesn’t agree with experiment, it’s wrong.” And he was a theoretician.

 

[Octopuss]

Less physics and e-peen, more human speech pls

 

[mtcn77]

Less physics and e-peen, more human speech pls

They call it a convex path, I'd call it orthogonal gyroscopic orbits. A picture is worth _ amount of words.

 

[agentnathan009]

You seem awfully insecure for someone without even a trace 'notion' of how electricity travels in the opposite path that electrons travel...
The issue is, I 'know' how electrons are even shaped like and they can lose track into unobserved paths since they are not shaped like a 'single' orbital physical ring, but double orbital rings with a second perpendicular one connected at the electron's moment center that can crossect without any discrimination for physical barriers - its quantum orbit does not interact with physical matter, just cuts across it. So much to say, electrons having an extra invisible quantum orbital ring that they can always switch between just as easily as the one they are physically travelling on - that is the very definition of Murphy's Law in electricity "what will go wrong, does" without you even knowing it.
Your understanding of electrons is just half of the story, you are trying to solve it as a solid matter. That is not how it intuitively behaves.
But as the famous physicist Richard Feynman once said, “It doesn’t matter how beautiful your theory is, it doesn’t matter how smart you are. If it doesn’t agree with experiment, it’s wrong.” And he was a theoretician.

I'm sorry for being rude and insulting to you.

I'm not insecure, only allowing for variances with other aspects that I don't understand as fully.

 

[ViperXTR]

OCCT 8.0 now has some test for Curve optimizer users to verify stability

https://twitter.com/i/web/status/1368847546278350850

 

[Octopuss]

Any idea how long should I keep OCCT tests running on each core? I'm used to prime95 where I won't settle for anything less than 12 hours without errors, but OCCT might do completely different things and this might not apply at all.

 

[mtcn77]

Any idea how long should I keep OCCT tests running on each core? I'm used to prime95 where I won't settle for anything less than 12 hours without errors, but OCCT might do completely different things and this might not apply at all.

Just run it until the test reaches your usual operating cpu temperature. Generally, I find no difficulty in setting the temperature just right, instead I find it hard keeping it down. The core will present an error in case its overclock is not stable. You can adjust the temperature using the FFT length gauge(speaking about P95). I set it 16-16 for a near original power draw that isn't inherently too synthetic.

 

[Octopuss]

Um, you know these CPUs reach the max temperature within about two seconds, right?

 

[mtcn77]

Um, you know these CPUs reach the max temperature within about two seconds, right?

Is it so hard to believe?

 

[Octopuss]

I have no idea what are you talking about.
I guess for you it is, otherwise you wouldn't suggest what you wrote (which didn't answer my question even partially).

 

[blu3dragon]

Any idea how long should I keep OCCT tests running on each core? I'm used to prime95 where I won't settle for anything less than 12 hours without errors, but OCCT might do completely different things and this might not apply at all.

I think it is similar to prime95 in the way it works. For a final stability test I would do at least one hour per core, but generally I've found prime95 to be better at finding errors than occt.

 

[Nantes]

In my BIOS for the B550-E Gaming you can reach "Precision Boost Overdrive" via two different routes: the "AI Tweaker" tab or or the "Advanced" tab. Both have nearly the same options except that the latter allows you to change the Curve Optimizer function while the former does not.

Upon further testing I have noticed that both PBOs behave vastly differently: turning the one under AI Tweaker on, my 5600x's power consumption shoots from 76W (stock draw) to 95W or more (it's also worth noting that turning on "AMD Performance Enhancer" under the same AI Tweaker menu has the same effect, and I've found no practical difference between enabling either that or PBO). Whereas turning on the PBO under the Advanced tab on with the same sub-options (and Curve Optimizer on Auto), power draw stays at 76W, the normal power draw constraint. It is clear that despite being named the same and having very similar sub-options, these two features are completely different.

Can someone knowledgeable please explain this mess?

Example images just to show the two different paths (the values for the options are not what I'm using):

 

[Zach_01]

PBO on auto means disabled.
Only on enabled is really enabled. That’s why you saw this difference.

 

[Mussels]

In my BIOS for the B550-E Gaming you can reach "Precision Boost Overdrive" via two different routes: the "AI Tweaker" tab or or the "Advanced" tab. Both have nearly the same options except that the latter allows you to change the Curve Optimizer function while the former does not.

Upon further testing I have noticed that both PBOs behave vastly differently: turning the one under AI Tweaker on, my 5600x's power consumption shoots from 76W (stock draw) to 95W or more (it's also worth noting that turning on "AMD Performance Enhancer" under the same AI Tweaker menu has the same effect, and I've found no practical difference between enabling either that or PBO). Whereas turning on the PBO under the Advanced tab on with the same sub-options (and Curve Optimizer on Auto), power draw stays at 76W, the normal power draw constraint. It is clear that despite being named the same and having very similar sub-options, these two features are completely different.

Can someone knowledgeable please explain this mess?

Example images just to show the two different paths (the values for the options are not what I'm using):

View attachment 198116

View attachment 198117

my board has this too, its weird to see two sets of the same settings that dont sync

 

[Nantes]

PBO on auto means disabled.
Only on enabled is really enabled. That’s why you saw this difference.

But see the second screenshot. In order to access all the options you can see, it must be set to "advanced", which must mean it is on, otherwise there would be no point in setting any of the options it reveals. Which of the OTHER options besides "PBO -> Advanced" must be turned on for PBO to work from this menu? PBO limits?

 

[Zach_01]

I apologize in advance as for some of the following I'm not entirely sure or cant remember which is which... Its been over a year since I was messing with OC settings and I'm old now...

1. I'm not entirely sure but if you choose "Advanced" mode you have to at least choose a "Max CPU boost clock Override" other than 0MHz. (fact=)Other settings can be on "Auto" or choose your own. If you just put it to Enable it does everything by it self.
2.(fact=) PBO Scalar when on Disabled/Auto is by default X1. Only on "Manual" and other than X1 is doing something. Be aware of it as past the X2-3 the voltage feed of the CPU could be too high, or even stupid high... (X4/5+)

3. Some/a lot settings are placed into 2 different places.
AMD CBS
AMD Overclocking

4. I think the one overrides the other but cant remember which one. Also settings on "AMD Overclocking" cant be returned to default on auto clear CMOS/default settings after several failed boot attempts that most boards do.

 

[529th]

Are rounding errors with SSE CoreCylcer test important?

 

[blu3dragon]

Are rounding errors with SSE CoreCylcer test important?

It means that your current settings are not stable. So yes, it is important.

 

[kiddagoat]

Ryzen 5000 Precision Boost Overdrive 2 Guide - YouTube

This video helped me out tremendously. He breaks this down and explains just about all of the settings. I found this one night after I got my 5900x.

I haven't dialed in my curve yet, but I am currently at -15 on all cores and I boost 4.7ghz all cores and I have seen single cores get up to 5.1-5.2ghz.

Lots of patience and trial and error.

 

[blu3dragon]

But see the second screenshot. In order to access all the options you can see, it must be set to "advanced", which must mean it is on, otherwise there would be no point in setting any of the options it reveals. Which of the OTHER options besides "PBO -> Advanced" must be turned on for PBO to work from this menu? PBO limits?

It might help to think of it like this:

PBO stands for precision boost override. Precision boost is the stock functionality that is used to calculate what clock speed to run the cpu at. It does this based on a number of factors including power, temp, voltage required for stability at a given frequency, max safe voltage, etc.

PBO allows you to over-ride some of the stock settings for power, temp, etc. Typically PBO is used to increase the stock power limits. You can also use it to lower them, as well as lower or raise temp limits.

So, when you "enable" PBO, you also need to make sure that the settings for power, temp, etc are as you want them. If they are all still at the default setting, then simply enabling PBO as a feature won't change anything.

In the ASUS bios, there is an ASUS provided menu for PBO control. This is the one under the "AI Tweaker" tab. When you enable PBO in this tab, the bios automatically increases the stock power limits. As far as I have seen, any settings for power limits, etc, in this tab take priority over ones in the Advanced menu. i.e. whatever power limit is set here will be the one that is used, regardless of what is set in the "Advanced" tab.

The "Advanced" tab has the options that AMD exposes as part of the standard AGESA implementation.

So, you basically have the ASUS menu (under AI Tweaker) and the AMD menu (under Advanced).

My recommendation is to do all of your settings under the ASUS provided "AI Tweaker" menu, since those settings take priority. Leave everything under Advanced->AMD Overclocking at default/auto or motherboard to make sure there is no conflict.
The one exception to this is when you come to tune curve offsets, since those are not available in the AI Tweaker section.
For those, adjust under Advanced->AMD Overclocking->Precision Boost Override->Curve Optimizer.

Be careful that if you disable PBO, the Curve Optimizer settings still take effect (even though the menu disappears in the bios). So, if you want to return to stock, you need to either reset the bios completely, or make sure to go in and zero all off the curve offsets before disabling PBO or returning any of the other settings to stock.

 

[Nantes]

I'm testing my curve optimizer undervolts with CoreCycler, and it turns out my new 5600x's Core 1 returns rounding errors even with no offset (negative offset set to 0). Does this mean this core is defective (and if so, is it worth it to activate my warranty just for this)? Or should I attempt to fix this by applying a positive offset?

@blu3dragon thank you for your info!