Explaining the AMD Ryzen "Power Reporting Deviation" -metric in HWiNFO

Alekcnito

New Member
MSIOverheating.png
Test: CINEBENCH R23(R20 is no longer in the microsoft store)
HWiNFO32 v6.40-4330

Motherboard: MSI b450-A PRO (BIOS Version:7B86vAD)
CPU: Ryzen 5 3600
Cooler: Cooler Master MASTERLIQUID ML240L V2 RGB
Case: Corsair Carbide 275R
GPU: evga 1660 ti
RAM: Corsair Vengeance LPX 16GB

The motherboard is at stock settings except for XMP.

I bought this motherboard a year ago and I’ve had temperature problems since I bought it. I didn’t know it was the motherboard and blamed the stock cooler for the extremely high temps(95° under load).

I installed a liquid cooler yesterday and nothing changed except for lower temperatures at idle.
I’ve tried reducing the voltage by offsetting to try to reduce the temperature but it only changed by 2°. I am now out of ideas on what to do other than buy a new motherboard(which I would prefer not to do). =’(

I keep thinking I’m destroying the CPU whenever I render a video on Premiere since It reaches the same temperature. I’m basically limited to gaming and lighter tasks until I fix the issue.

Is there anything else I should try before buying a new motherboard? Is the problem not the motherboard? Can someone help please.

Motherboard
Motherboard.jpg
Cooler
https://www.coolermaster.com/la/es-la/catalog/coolers/cpu-liquid-coolers/masterliquid-ml240l-v2-rgb/
Cooler.jpg
 

Zach

Well-Known Member
View attachment 5679
Test: CINEBENCH R23(R20 is no longer in the microsoft store)
HWiNFO32 v6.40-4330

Motherboard: MSI b450-A PRO (BIOS Version:7B86vAD)
CPU: Ryzen 5 3600
Cooler: Cooler Master MASTERLIQUID ML240L V2 RGB
Case: Corsair Carbide 275R
GPU: evga 1660 ti
RAM: Corsair Vengeance LPX 16GB

The motherboard is at stock settings except for XMP.

I bought this motherboard a year ago and I’ve had temperature problems since I bought it. I didn’t know it was the motherboard and blamed the stock cooler for the extremely high temps(95° under load).

I installed a liquid cooler yesterday and nothing changed except for lower temperatures at idle.
I’ve tried reducing the voltage by offsetting to try to reduce the temperature but it only changed by 2°. I am now out of ideas on what to do other than buy a new motherboard(which I would prefer not to do). =’(

I keep thinking I’m destroying the CPU whenever I render a video on Premiere since It reaches the same temperature. I’m basically limited to gaming and lighter tasks until I fix the issue.

Is there anything else I should try before buying a new motherboard? Is the problem not the motherboard? Can someone help please.

Motherboard
View attachment 5680
Cooler
https://www.coolermaster.com/la/es-la/catalog/coolers/cpu-liquid-coolers/masterliquid-ml240l-v2-rgb/
View attachment 5681
I asume this sceenshot is during 100% load right?

First of all, the motherboard is lying about CPU consumption. But i dont think this is your problem, and let me explain.

See on your screenshot the current "CPU PPT" value reporting 58,5Watt. Having a current PowerReportingDeviation at 85% means that the true power consumption of the CPU (at that exact time) is 58,5 / 0.85 = 68.8W. With a power around 69W a CPU should never reach 95C and throttle if the cooling is sufficient. You can see that its throttling by the current value of "CPU core voltage (SVI2 TFN)" beeing only at 0.975V. I bet frequency was also reduced at that time.

Take a look at mine on Cinebench R20 right before it ends:

HWiNFO_30_12_2020_b.png

See the CPU PPT at 87W and PRD at 91%.
So 87 / 0.85 = 102W. Check also the CPU voltage (SVI2 TFN) and the core clock. I red boxed everything you need to see.
Im using a 280mm AIO and the fans are only 1200~1300rpm. Ok, Im not using a case and the ambient around the AIO radiator is about 24C but still, you shouldn't have that level (95C) of temperature. Something is either wrong with the mounting mechanism of the CPU socket or the thermal paste, or the airflow of the case. Though I doubt about the air flow as all other temps are normal.

Did you ever update the board BIOS to the latest version?
 

Alekcnito

New Member
The screenshot is at 100% load.

I have removed and reseated the cooler multiple times and finally decided to change it to a liquid cooler. The temperature readings are the same so I don't think its an improper installation on my part. Could it be a problem with the motherboard?

I updated the BIOS earlier this year but haven't installed the latest version. I will try that today.

Thank you Zach. Your insight has been very helpful in understanding the problem.
 

Azaghoth

New Member
Just found out about this and ran a few tests on R20 for 10 minutes, saw deviation measured at 84.5% the entire time with a 100% load under R20. Using a Corsair H100i on an AMD Ryzen 7 3700x /w Gigabyte Aorus I Pro Wifi, CPU was averaging about 4.15ghz per core. Nothing but PBO on for the CPU in BIOS and everything else on auto, max temps under full load reach 69c(nice). Voltages average 1.323 and maxed at 1.475, while CPU package wattage averaged around 104W, but the CPU PPT limit was on average 15%?? This was with voltage settings in BIOS at auto, once set PBO off, along with setting all core voltages to normal, I saw the clocks boost to 4.23ghz and max out at 61c on the hottest core, while the deviation remained the same. However, on this run I saw core wattage average about 70w and the CPU PPT was averaging 85%. R20 scored 7931 on the latter settings while it scored 12414 with auto on voltages and PBO on. What does this mean exactly?
 
Last edited:

Zach

Well-Known Member
Your CPU has the following PB (Precision Boost) limits

CPU PPT: 88W
CPU EDC: 90A
CPU TDC: 60A

When you have "PBO: Enabled" in BIOS the system is expanding (overrides) those limits and the CPU whenever it can may cross the above limits acccording to internal parameters (if they allow it). You can see what those (new) limits are in RyzenMaster's Home page >> Control Mode
PBO: Precision Boost Override

You are refering to a lot of different sensors. "CPU PPT" and "CPU Package Power (SMU)" should be around the same at all times (on any load). If they are more than 1~2W different I'd stick with PPT. "CPU Core Power" is only the core power and does not represent anything useful.
PPT: Package Power Tracking

Power Reporting Deviation of 84.5% means that the board is understating (feedback info) the total CPU power by around 18% to the CPU it self.
So the true PPT is "PPTvalue divided by 0.845"

Have the latest stable BIOS for your board, the latest chipset drivers from AMD web page directly and latest Win10 version (20H2).
 
Last edited:

maflynn

Member
I'm not sure what to make of his, I just googling his after installing HWINFO onto my freshly built system.

I'm using:
Ryzen 3700x
Gigbyte X570 i Aorus Pro Wifi
The left most exhibit is at idle, and the right is under load. Using the equation to compute actual wattage (PPT / Power Reporting Deviation) I get the following
Idle: (88 / 25.3%) = 247 watts
Load (88 / 83.1%) = 105 watts

The underload value seems more likely, but even then my temps are so low (45.3c) and I've run this multiple times. I really don't think the reported metric is all that accurate, is my 3700x really pulling in 247 watts with a temp of 33?


1611771335727.png
 

Zach

Well-Known Member
No of course not.
It has been said many times here and also in the very first post of this topic, by the man who found this... that PRD(%) is valuable only on 100% CPU load or near 100% load with stock settings for CPU power management. So for idle or even gaming loads its not making any sense and it shouldn’t be a concern.
 

Zach

Well-Known Member
As for the calculation you don’t use the default PPT, but what the board is reporting during the run.

Also... what’s up with that super low UCLK/FCLK speeds?
What’s your RAM speed?
 

maflynn

Member
It has been said many times here and also in the very first post of this topic, by the man who found this... that PRD(%) is valuable only on 100% CPU load or near 100% load with stock settings for CPU power management. So for idle or even gaming loads its not making any sense and it shouldn’t be a concern.
Fair enough, but at this point it seems that perhaps the developers update the program to display accurate information.

so... what’s up with that super low UCLK/FCLK speeds?
No idea, I didn't even notice that. I just enabled XMP, so its now 1,600 for both. That value was coming from a stock bios setting - not sure why it was so low.

My memory is Corsair Vengeance LPX 32GB (2 x 16GB) DDR4-3200 PC4-25600 CL16 Dual Channel
 

Martin

HWiNFO Author
Staff member
There's nothing to fix. This feature i meant to be used at full load only and can provide reliable results only in such scenario.
 

maflynn

Member
There's nothing to fix. This feature i meant to be used at full load only and can provide reliable results only in such scenario.
Then don't display anything until the system is at full load. As it stands when the system is not at full load, its displaying inaccurate numbers, pure and simple. There's nothing in app to communicate that the number being exhibited is not correct under normal non-stress test or full load conditions
 

Zach

Well-Known Member
That’s why this topic exists and created by the one that found this issue with boards that lying about power consumption. To expose it and explain the mechanics behind this new value. Like a lot sensors in HWiNFO that is new, and users are unfamiliar with. Like “Effective Clock” for another example. Do you see any description about what it is exactly anywhere into the app itself? Here in the forum exists.

By your calculations

Idle: (88 / 25.3%) = 247 watts
Load (88 / 83.1%) = 105 watts


...its obvious that you didn’t take the time to read the first post of this very topic. As I said earlier you don’t divide the stock rated PPT of the CPU with the PRD value.
You divide the reported PPT value with the reported PRD value during the 100% load.
 

Martin

HWiNFO Author
Staff member
Then don't display anything until the system is at full load. As it stands when the system is not at full load, its displaying inaccurate numbers, pure and simple. There's nothing in app to communicate that the number being exhibited is not correct under normal non-stress test or full load conditions

This has been under consideration, but it's not always possible to determine based on factors like "Total CPU Usage" whether the system is currently under sufficient load for this purpose.
 

maflynn

Member
...its obvious that you didn’t take the time to read the first post of this very topic. As I said earlier you don’t divide the stock rated PPT of the CPU with the PRD value.
You divide the reported PPT value with the reported PRD value during the 100% load.
You completely missed my point. I concede that I did not read the entire first post, but that's the issue. Why should I spend so much time googling and researching about this one metric and then reading the post in a forum simply because its displaying an inaccurate value. No matter how you slice it, what is being displayed is inaccurate. You're providing a reason for such inaccuracy, but why should I have to spend time researching that when in my opinion an application should only display accurate values.

I'm new here, and I'll not look to ruffle feathers, I got the answer I needed, but let me close and say that if an application (any application) is displaying a value, then a typical user will read that at face value. The fact that this thread is 17 pages is testament to the confusing nature (and no I did not read all 333 posts).

I appreciate the hard work the developer put into this application, is a great application but I think if any application displays a value that is incorrect/misleading because certain situations have not been achieved , then that's a mistake. Its my opinion, and I stand by that. I appreciate the fast responses to my query, and I hope everyone does not take offense to my suggestions.

No need to beat a dead horse at this point as I said, I have what I need
 

Zach

Well-Known Member
I dont think that anyone is or should be offended. You have your opinion and you are entitled to one. I strongly disagree with yours and thats mine.
I didnt miss your point at all. I can understand how confusing are such matters to most users who just want to use their system and come across this. There is no need to read 300+ posts. Just the first would suffice. If one hasnt read it cant understand the nature of the matter. Still many didnt get it after reading it. Its not all straight forward IMHO. The report of PRD value is only valueble to those who want to know their system better and how their board is treating their CPU. They want to have it, and others dont care.
I also believe(or even convinced) that because of this gone wide through HWiNFO, a lot of board vendors had their BIOS straightened out and stop "cheating" over competion.

Even if it was possible to report an accurate value in a non 100% load scenario it would be useless, irrelevant and insignificant because this has true value when the CPU is on the limit of multithreading. If that limit set by AMD is crossed or not without the user's intention and awareness.
 

cbell4

New Member
Posted for reference. On my ASRock X399 Taichi MB with Ryzen 1950X running CineBench R23, the PRD was 102.2%. Nice to know that, for this MB at least, one manufacturer is being honest by reporting actual values that don't deviate from expected values in a statistically significant way.
 

AeroZ

Member
I noticed it before that the PRD value was red. Googled it and landed here. Gigabyte reporting is pretty off.
 

Attachments

  • PRD.png
    PRD.png
    99.7 KB · Views: 6
Last edited:

Zach

Well-Known Member
Get the latest beta of HWiNFO. Has a tooltip function on all sensors. This value has no meaning unless CPU load is near or at 100%.
 

Zach

Well-Known Member
In that case yes your board is very off in reporting Power. Even so, the CPU does not exceed its limits. Your CPU has 125W PPT limit, and on your screenshot we see a true power of just 100W (67 / 0.67 = 100)
Its not Gigabyte specific but maybe BIOS version specific. My X570 AorusPro has a 91~94% PRD.

This is under R20
PPT reports 85W with 92% PRD (85 / 0.92 = 92W)
My 3600 has a 88W PPT limit but because of low temp the CPU allow itself to go a little over 88W.

1613493371220.png
 
Top