Confusing ZEN 2 voltages

[Dux]

Idk what is going on with reported voltage of my R5 3600 CPU. When i boot to BIOS, it says CPU voltage is 1.460V. In windows CPU-Z reports 1.408V, OCCT reports 1.41V, Ryzen clock tuner reports 1.250V, and Ryzen master reports 1.3V current and maximum because that's what i manually set for my OC to 4.2GHz. So which value is correct?

 

[Zach_01]

Skip these monitoring tools except RyzenMaster.
If you want a real monitoring tool for the entire system you will go for HWiNFO.

Open in "Sensor only mode" and watch the "CPU Core voltage (SVI2 TFN)"

Second row, somewhere in the middle, see the current, min, max, avg values:



Leave it open in systray and use your system.
Click the gear down right to open settings and set poling period from 2000ms to 1000ms, click SET and OK. This will make it more accurate, and it doesnt load the system

You will see differences with RyzenMaster but thats normal. RyzenMaster uses AMD proprietary methods for all values (voltages, speeds, temps) that do not disclosure and so no one knows.

 

[mtcn77]

Very simply, if you didn't set the 'scalar' item [Auto], or 2x at most, your cpu will always spike above safe limits.

 

[Zach_01]

That wont work, he is OCing.
The question was about discrepancies among the "tools" for voltage.

We will talk about that OC after...

 

[Dux]

Well, i did set to auto OC now in ryzen master. Voltage actually dropped to 1.1V in Ryzen master and frequency jumped from 4.1 to 4.2GHZ. Tested with OCCT and stable, no errors. Maintains constant 4.2GHz. Tested in Cinebench 20. Went from 3644 to 3839 score.
Hwinfo shows this


Can't go beyond 4.2GHz no matter the voltage. What is the reason for this limit? Why are ZEN processors such bad overclockers compared to intel? What if i raise FSB in bios and add some frequency that way? Yes i know that also OC's RAM, but i can compensate for that.

 

[mtcn77]

Go to scalar, it has to be auto(1x), or 2x.
Afterwards, don't touch anything and try your hand with pbo. What are the baselines?

1. PPT: 88W,
2. TDC: 60A,
3. EDC: 90A.

First thing is first, ppt has to be 10% higher than edc. Then, if you want boost edc has to increase. Try in increments of '5'. If your voltage is set low, you can play around with options without triggering the TDC throttle. It really doesn't do much, 1.4v is 84w whereas 1.25v is 75w - you won't be able to make a breakthrough at 84 watts from what you couldn't do in 75 watts, so don't use scalar, and do what you can sticking with the onboard monitoring of the chip.

 

[Zach_01]

Well, i did set to auto OC now in ryzen master. Voltage actually dropped to 1.1V in Ryzen master and frequency jumped from 4.1 to 4.2GHZ. Tested with OCCT and stable, no errors. Maintains constant 4.2GHz. Tested in Cinebench 20. Went from 3644 to 3839 score.
Hwinfo shows this
View attachment 172154

Can't go beyond 4.2GHz no matter the voltage. What is the reason for this limit? Why are ZEN processors such bad overclockers compared to intel? What if i raise FSB in bios and add some frequency that way? Yes i know that also OC's RAM, but i can compensate for that.

CPU EDC (~112A) is far beyond its limits of 90A with "only" 90W PPT.

Was that EDC (112A) under full load?

This level of EDC I would expect to see at 120~130W PPT at least, if not more.
If that wasnt full load, then I dont want to see how high this value goes under constant full load.

R5 3600 is not ment to work 4.2GHz all cores with 1.35V. You are stressing too much the silicon in my opinion, and I would expect a degradation at some point.
ZEN2 CPUs are pushed close to their limits already by AMD and really dont have much more to give in terms of performance. You can only try to adjust efficiency and that isnt certain to success either.

You cant overclock with these systems today with BaseClock (the old FSB). If you do that you also oveclocking the Chipset and PCI-E links and that affects stability of the system. You may risk data corruption on the M.2 drive, along with other issues/causes of instability.

I think you should backdown a little the voltage and frequency if necessary.

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

If you want to determine the max voltage that the CPU can take under load you should turn everything on Auto and run Prime95 with custom FFTs 128K and see the voltage that power management is feeding the CPU.
This is only an example:
If power management feeds the CPU with 1.25V under Prime95 128K FFTs, then you should set voltage to around 1.3V and apply some LLC Vdroop so under heavy load the voltage to drop to 1.25V. And of course adjust the frequency to that voltage.

 

[mtcn77]

CPU EDC (~112A) is far beyond its limits of 90A with "only" 90W PPT.

Was that EDC (112A) under full load?

This level of EDC I would expect to see at 120~130W PPT at least, if not more.
If that wasnt full load, then I dont want to see how high this value goes under constant full load.

R5 3600 is not ment to work 4.2GHz all cores with 1.35V.

That is the point with EDC. EDC>PPT give single thread boosting the green light.
It is so very intricately balanced, you see... every line of the program solves a vital portion of the puzzle with none of the parts left out of the picture.

• EDC>PPT(base): meaning, "I want to boost single threadedly as frequently as possible",
• PPT>EDC(base): meaning, "I want boost at multithreaded loads",
• EDC<PPT(base): meaning, "I want no boost, but cores working equally".

 

[Zach_01]

And that applies to auto power management settings or under static ones?
Because when you do static speeds and voltages the power management has its hands tied. And the CPU(silicon) is on mercy of load. I think?

EDIT:
typo

 

[Dux]

CPU EDC (~112A) is far beyond its limits of 90A with "only" 90W PPT.

Was that EDC (112A) under full load?

This level of EDC I would expect to see at 120~130W PPT at least, if not more.
If that wasnt full load, then I dont want to see how high this value goes under constant full load.

R5 3600 is not ment to work 4.2GHz all cores with 1.35V. You are stressing too much the silicon in my opinion, and I would expect a degradation at some point.
ZEN2 CPUs are pushed close to their limits already by AMD and really dont have much more to give in terms of performance. You can only try to adjust efficiency and that isnt certain to success either.

You cant overclock with these systems today with BaseClock (the old FSB). If you do that you also oveclocking the Chipset and PCI-E links and that affects stability of the system. You may risk data corruption on the M.2 drive, along with other issues/causes of instability.

I think you should backdown a little the voltage and frequency if necessary.

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

If you want to determine the max voltage that the CPU can take under load you should turn everything on Auto and run Prime95 with custom FFTs 128K and see the voltage that power management is feeding the CPU.
This is only an example:
If power management feeds the CPU with 1.25V under Prime95 128K FFTs, then you should set voltage to around 1.3V and apply some LLC Vdroop so under heavy load the voltage to drop to 1.25V. And of course adjust the frequency to that voltage.

Dude, i literally posted to you what you asked from Hwinfo, and looked under voltage where you said. And now you say completely the opposite. Both Ryzen master and Hwinfo report 1.1V core voltage. I don't think you yourself understand what a bunch of these values you are mentioning actually mean.

 

[mtcn77]

If you want to determine the max voltage that the CPU can take under load you should turn everything on Auto

Just a pointer, if you do that, you reset to "the motherboard auto" levels which are different than manufacturer original which are called "manual", ironically. Auto involves OEM overclocking, they do this to top the charts in motherboard reviews.