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

Dalai

Well-Known Member
Despite the 3070FE running cooler overall, it's a dual fan card that dumps its heat into the case. I did not think it would affect temps THIS much [...]
While this is certainly true, it doesn't matter at all when you run a CPU stress test like Prime95, CPU-Z's benchmark, CineBench or similar. The GPU doesn't do anything during these tests. When the CPU hits its max temp this fast, you have a CPU cooling issue.

Regards
Dalai
 

Steven409

Member
While this is certainly true, it doesn't matter at all when you run a CPU stress test like Prime95, CPU-Z's benchmark, CineBench or similar. The GPU doesn't do anything during these tests. When the CPU hits its max temp this fast, you have a CPU cooling issue.

Regards
Dalai
So you think it would be a good idea to remove cooler and reapply thermal paste and/or replace cooler? Or, let it be since temps are still within acceptable range outside of stress testing? I forgot this CPU cooler on this particular system is not the stock cooler, it's actually a Cooler Master.

I'm having a hard time figuring out the PRD setting and why it is reporting <50% under full load. My understanding is the only way to correct this is through a BIOS update and I've tried to find a BIOS update for this particular motherboard, and think I've found it, but cannot be entirely sure since it's not listed on ASrock's website.

The reason I've never looked to update the BIOS before is because I've never had an issue...it's worked just fine for over 2.5 years until just a few days ago when I changed the GPU. I'm trying to think if I hit anything or move anything around inside the case and I've inspected it several times over.

Would a bad cooler cause the PRD readings to be this off?
 
Last edited:

Zach

Well-Known Member
EDIT:
Something else I noticed with this new version of HWiNFO is that it has a new temp reading of "CPU Die (average)". The last version I had did not have this and I've always used the CPU CCD1 reading. Is the CPU Die a better reading to go off of?

When the CCD1 reading shows current max of 72C, the CPU Die shows 65C. The CPU Die reading seems to average about 3-5C less.
So you think it would be a good idea to remove cooler and reapply thermal paste and/or replace cooler? Or, let it be since temps are still within acceptable range outside of stress testing? I forgot this CPU cooler on this particular system is not the stock cooler, it's actually a Cooler Master.

I'm having a hard time figuring out the PRD setting and why it is reporting <50% under full load. My understanding is the only way to correct this is through a BIOS update and I've tried to find a BIOS update for this particular motherboard, and think I've found it, but cannot be entirely sure since it's not listed on ASrock's website.

The reason I've never looked to update the BIOS before is because I've never had an issue...it's worked just fine for over 2.5 years until just a few days ago when I changed the GPU. I'm trying to think if I hit anything or move anything around inside the case and I've inspected it several times over.

Would a bad cooler cause the PRD readings to be this off?
As I said already, your CPU under stress test is consuming more than 120W with a stock cooler for a CPU that its meant to draw around 90W max... This is a board doing.
We dont really know the info from before with 5700 about the CPU power and the power reporting deviation.

I do not know why this happened after a GPU change, but if everything else is the same I dont see how the CPU cooler gone bad all of a sudden either.
And yes during simple tasks or even a CPU stress test the GPU is not generating any significant heat so its not it.

Now taking account the spikes you are talking about at idle and the increased average temp overall I'm thinking about windows power plans as you mentioned before. See what you're using now and try to change it maybe. A power plan can make the CPU be much more aggressive and responsive to any kind of load and spike its boost.

And you really have to check if ASRock has a latest BIOS for your board that you dont have. And ALWAYS download BIOS files from manufacturer website for your board.
What do you mean by >> you think you found one but its not on ASRock's website?

As for the different temp sensors on HWiNFO, it goes like this:

_CPU (Tctl/Tdie) = Its the hotspot of the entire CPU. The reading is switching almost instantly to report the highest temp of all temp sensors into the CPU. A single CCD has more than 50 temp sensors onboard and this reports always the highest. Also this is the FAN cooler control temp (aka Tctl). This one spikes the most and sometimes makes the FAN rpm to spike too. Annoying to some users. Introducing a few seconds latency (delta) can make the cooler FAN ignore the spikes and work "smoother".

_CPU Die (average) = Its the average temp of all CCDs (1 or 2) depending the CPU. The average reading of all temp sensors inside CCD(s).

_CPU CCD1 (Tdie) = This reports all times a specific location sensor on one side of the CCD. If the CPU has 2 CCDs there is a second reading (CPU CCD2).

The CPU(Tctl/Tdie) should be the highest reading of the above 3, almost all times as this is its job (hotspot).
Between the other 2 depends on the type of load and the amount of load (%), but both should be fairly close to each other with maybe the CCD1 (Tdie) to be the highest (depending on what side of CCD is loaded)

There is not right or wrong or to what temp a user should be looking. Every reading reports a different aspect.
At 100% CPU load (stress test) all 3 readings should be very close to each other.
 
D

Deleted member 15652

Guest
its not on ASRock's website
He meant this motherboard doesn't officially exist. No such board on their website. What could be unclear here?
 
Last edited by a moderator:

Dalai

Well-Known Member
So you think it would be a good idea to remove cooler and reapply thermal paste and/or replace cooler?
Definitely check the cooler, as I said above. Check if it's mounted properly, and if it's warm to the touch. If you find either of those to be not the case, remount the cooler. Reapplying thermal paste is not a must if there's enough paste left, and it doesn't really matter for testing. But if there's not enough paste between the cooler and the CPU, then applying new paste is a good idea.

Or, let it be since temps are still within acceptable range outside of stress testing?
I don't consider temperatures above 50°C while using a web browser acceptable.

I forgot this CPU cooler on this particular system is not the stock cooler, it's actually a Cooler Master.
Well, it's probably better than the stock cooler, but maybe not. It doesn't matter which particular cooler you have if it's not mounted properly, or if it isn't capable of cooling the CPU ;).



The Asrock B550AM Gaming is an OEM board using the B550A chipset (which is different from the B550 chipset, IIRC). GamersNexus made a video about this chipset, maybe even about this particular board, some time ago.

Regards
Dalai
 

Steven409

Member
Thank you Zach, for the detailed descriptions of what each reading does. That helped more than even the pop-up window in the program itself and each one is very close during a stress test.

This build is actually a CyberPowerPC that I got at close out a couple years ago...paid less for the entire system than what it would've cost me to buy a 5700 and 3700X alone so I could not pass it up. I know a lot of people have nothing but bad things to say about them, but I've owned two and never had an issue until now. I may be making my "issue" bigger than it needs to be, though.

With that said, neither ASRock nor Cyberpower officially have the BM550AM Gaming motherboard or a BIOS listed on their websites. I did find this directory after searching other forums, but the fact it's not a secured site and 2 of the 3 file labels are not the proper name, I'm a bit skeptical (http://www.cyberpowerinc.com/drivers/?dir=Motherboards/MB-478-101 B550AM GAMING/BIOS).

I switched the Windows Power Option to Ryzen Balanced. It was on Windows High Performance, which is what I thought I had it set at previously, but now I'm not so sure. If it got changed, it was the recent Windows Update. The idle temps seem to be better with far fewer spikes, but the temps still rise quickly on both CPU-Z and Cinebench. The temps while using the browser is now in the low-mid 40s with a few spikes to upper 50s.

I just tested it out in a few games with Control and Cyberpunk being the two that got the hottest after 20minutes. Even then, Control recorded an average of 69.1C and Cyberpunk an average of 72.0C. Both games were at 1440p with the highest settings and all ray tracing effects turned on. The max temp recorded during both games was 79C and that was just on a quick spike here and there according to the graph. The temps seemed to stay pretty steady during both games and there was no obvious climb...they pretty much stopped at what the average temp was reported as.

Dalai, I checked the cooler again and even tried to tighten the screws but they were already tight. Considering the extra heat in the case from the GPU (max temp 68C), and the extra processing power used for the extra effects such as ray tracing, these temps are not too far off from what I had before.
 

Zach

Well-Known Member
Thank you Zach, for the detailed descriptions of what each reading does. That helped more than even the pop-up window in the program itself and each one is very close during a stress test.

This build is actually a CyberPowerPC that I got at close out a couple years ago...paid less for the entire system than what it would've cost me to buy a 5700 and 3700X alone so I could not pass it up. I know a lot of people have nothing but bad things to say about them, but I've owned two and never had an issue until now. I may be making my "issue" bigger than it needs to be, though.

With that said, neither ASRock nor Cyberpower officially have the BM550AM Gaming motherboard or a BIOS listed on their websites. I did find this directory after searching other forums, but the fact it's not a secured site and 2 of the 3 file labels are not the proper name, I'm a bit skeptical (http://www.cyberpowerinc.com/drivers/?dir=Motherboards/MB-478-101 B550AM GAMING/BIOS).

I switched the Windows Power Option to Ryzen Balanced. It was on Windows High Performance, which is what I thought I had it set at previously, but now I'm not so sure. If it got changed, it was the recent Windows Update. The idle temps seem to be better with far fewer spikes, but the temps still rise quickly on both CPU-Z and Cinebench. The temps while using the browser is now in the low-mid 40s with a few spikes to upper 50s.

I just tested it out in a few games with Control and Cyberpunk being the two that got the hottest after 20minutes. Even then, Control recorded an average of 69.1C and Cyberpunk an average of 72.0C. Both games were at 1440p with the highest settings and all ray tracing effects turned on. The max temp recorded during both games was 79C and that was just on a quick spike here and there according to the graph. The temps seemed to stay pretty steady during both games and there was no obvious climb...they pretty much stopped at what the average temp was reported as.

Dalai, I checked the cooler again and even tried to tighten the screws but they were already tight. Considering the extra heat in the case from the GPU (max temp 68C), and the extra processing power used for the extra effects such as ray tracing, these temps are not too far off from what I had before.
I agree that Ryzen Balanced plan is maybe the best way to go.

As for the loaded situations.. well you have options.
You can leave it as it is and be happy about it since the max temps on your regular tasks is no where near the max operating temp (95°C), (stress tested aside)...
...or you can check the BIOS for a specific setting that can bring PowerReportingDeviation closer to 100% and to the CPU intended PPT(88W), but with an expected loss in performance.

If you manage to drop power from 125+W to lets say 100W, temps will improve significantly because of the -20% decrease in power. This does not mean you will loose 20% of performance.
Performance and power is far from being linear to each other. Its a curve that gets steep towards the right end. I wouldn't be surprised if the performance loss is 5%, or even less at 100% load.

This can affect all kinds of workloads proportionally and not only the 95~100%, in terms of power and performance and temps.

The actual settings in BIOS that can adjust the power of the CPU closer to its default setting is the
"CPU VDD Full Scale Current" and "CPU VDD Telemetry Offset Value"
The values are in A (ampere) or mA.

Unfortunately I cant help more as to what setting you should input there. All I can tell is that those with <100% PRD are dialing a negative value and those with >100% PRD a positive one.
You will have to search more the web for info and dont do anything until you're certain.
 

Steven409

Member
I agree that Ryzen Balanced plan is maybe the best way to go.

As for the loaded situations.. well you have options.
You can leave it as it is and be happy about it since the max temps on your regular tasks is no where near the max operating temp (95°C), (stress tested aside)...
...or you can check the BIOS for a specific setting that can bring PowerReportingDeviation closer to 100% and to the CPU intended PPT(88W), but with an expected loss in performance.

If you manage to drop power from 125+W to lets say 100W, temps will improve significantly because of the -20% decrease in power. This does not mean you will loose 20% of performance.
Performance and power is far from being linear to each other. Its a curve that gets steep towards the right end. I wouldn't be surprised if the performance loss is 5%, or even less at 100% load.

This can affect all kinds of workloads proportionally and not only the 95~100%, in terms of power and performance and temps.

The actual settings in BIOS that can adjust the power of the CPU closer to its default setting is the
"CPU VDD Full Scale Current" and "CPU VDD Telemetry Offset Value"
The values are in A (ampere) or mA.

Unfortunately I cant help more as to what setting you should input there. All I can tell is that those with <100% PRD are dialing a negative value and those with >100% PRD a positive one.
You will have to search more the web for info and dont do anything until you're certain.
Thank you (and everyone else) for their input. I honestly don't think I would've ever posted here or been as concerned if I had not updated HWiNFO and saw that PRD setting and the readings that reported it to be off. I've since found a GamerNexus video that mentions this specific thread and even shamed ASRock as being one of manufacturers who fool the motherboard into drawing more power for the CPU so they'll look better on review builds.

I've never been one to OC my systems and so I never really play around with the voltage and power of CPUs or really much of anything in the BIOS. I'm not one who needs to have 315fps in games and I'm perfectly happy with a steady 60fps which has always been more than capable of achieving without the need to OC. Even before, I would just lower settings to get that 60fps. I just thought I may have screwed something up unintentionally while swapping the GPU despite doing that many times in the past with other builds.

Since the idle temps are more within acceptable range after changing the Windows Power Options, in addition to Cyberpunk, Control, and Red Dead 2 (three of the most graphically demanding games I have) averaging 68-72C with all settings maxed out and never getting above 79C, I think I'm going to leave it as is. That's all with an ambient room temp of 27.2C (81F) and still achieving 70-80fps average.

With everything else working fine, and this motherboard not "officially existing" according to ASRock, I don't want to take the chance of updating the BIOS and screwing something else up. I intend on the system lasting me at least another year or 2.
 

Steven409

Member
Just in case anyone comes here like I did with a similar problem looking for a solution, know that my temps kept getting worse so I decided to replace the CPU cooler completely. In removing the old cooler, I noticed the thermal paste was extremely poor and almost non-existent in the center of the IHS which would explain the dramatic jumps in temps I was getting. It was about time to reapply the paste anyway, but even more so considering the crappy application job it had upon being built. Figured I might as well replace the cooler with a better one while I had it off.

With the new cooler (CoolerMaster Hyper 212 Black Edition) and new paste (Arctic MX-4) applied, I'm now averaging 33C at idle in a 28C room. Max temps in gaming have yet to go over 65C and using Cinebench, it averaged only 77C after 5 minutes of running and never got above 80C (it was previously hitting 90C after just 10 seconds). This new cooler is so, so much quieter as well.
 

contol

New Member
There is typically a delta between the software-reported values and measurements taken at the physical layer (i.e., the 8-pin) due to VRM efficiencies. Assuming the software is reporting correctly, and that the settings are correct in the BIOS, what type of delta should one expect between measurements at the eight-pin and the software-reported values? 8-10%?
 

Zach

Well-Known Member
There is typically a delta between the software-reported values and measurements taken at the physical layer (i.e., the 8-pin) due to VRM efficiencies. Assuming the software is reporting correctly, and that the settings are correct in the BIOS, what type of delta should one expect between measurements at the eight-pin and the software-reported values? 8-10%?
I believe this can vary by a lot. As you said its due to VRM efficiency. But different boards use different (VRM) modules that each have their own curve of efficiency. The same VRM has different eff at 10A, different at 20A, 30A... and so on. Unless you have the data for a (your) specific module(s) you cant find out anything. And even so you cant really know how the board is treating them. When a board has a lot (VRM) phases it may shut down a few when the power requirements of the CPU is low compared to the capabilities of the power delivery system as a whole.

Its too complicated me thinks.
 

Haseo

New Member
This is a bit over my head, but I let HWinfo run for like 6 hours on my AMD Ryzen 5 2600 paired with an ASUS Prime B450M-A/CSM motherboard. The power reporting deviation was as low as 81.2% and as high as 152.2%... i'm assuming this is not normal? I did not run any benchmarks during that time. Simply generic usage and some gaming.
Also, the CPU is running at stock voltages and speeds with overlocked RAM (3200MHz clocked down slightly to 3133MHz due to some odd stability issues).
 

Attachments

  • CPU_numbers.png
    CPU_numbers.png
    61 KB · Views: 5

Zach

Well-Known Member
This is a bit over my head, but I let HWinfo run for like 6 hours on my AMD Ryzen 5 2600 paired with an ASUS Prime B450M-A/CSM motherboard. The power reporting deviation was as low as 81.2% and as high as 152.2%... i'm assuming this is not normal? I did not run any benchmarks during that time. Simply generic usage and some gaming.
Also, the CPU is running at stock voltages and speeds with overlocked RAM (3200MHz clocked down slightly to 3133MHz due to some odd stability issues).
Its very normal.
Power reporting deviation only matters under 100% CPU load. You can find more info on the very first post on this thread and on info tooltip if you hover your pointer over PRD sensor
 

prezz

New Member
I need a hero because I have no idea what is happening. I have replaced my CPU to a 5900x from a 3700x, and the deviation is super low when I run r23. It doesn't go above 60 on full load. The processor's score in r23 is only half of what it should be normally, around 11000 (should be 22000).
I also had a fresh install of windows on another nvme drive in the same computer, so I booted it up and ran the same tests. Perfect deviation, doesn't go below 93 on full load, temps are overall down, and the performance of the CPU on r23 is perfectly normal (twice the reported score of the above, around 22000 on Balanced power plan).

I thought it was maybe some power settings that I changed, so I changed everything to balanced and no dice. It's literally the same PC, just two different nvme drives and two installs of windows running latest chipset drivers and the like. I am hitting a brick wall here, what the heck is going on?
 

Attachments

  • Test Fresh System.jpg
    Test Fresh System.jpg
    501.6 KB · Views: 7
  • Test Old System.jpg
    Test Old System.jpg
    439.5 KB · Views: 7

Dalai

Well-Known Member
@prezz:
It looks like you missed the obvious thing: One of the Windows installs (Old System) is recognizing the 5900X as a six-core CPU with 12 threads. What could be causing this? Not sure. Maybe you have Ryzen Master installed and set it to Game Mode (or whatever it's called) which disables half the cores. I remember such a setting having positive effects on Threadripper systems in their early days. But it could be other things causing this. One thing is certain: Half the cores (or SMT) are/is disabled and that's why you see such behavior and half the scores in CineBench.

Regards
Dalai
 

prezz

New Member
@prezz:
It looks like you missed the obvious thing: One of the Windows installs (Old System) is recognizing the 5900X as a six-core CPU with 12 threads. What could be causing this? Not sure. Maybe you have Ryzen Master installed and set it to Game Mode (or whatever it's called) which disables half the cores. I remember such a setting having positive effects on Threadripper systems in their early days. But it could be other things causing this. One thing is certain: Half the cores (or SMT) are/is disabled and that's why you see such behavior and half the scores in CineBench.

Regards
Dalai
That is incredibly helpful. I see now that the average active core count is only 6. I think I messed with some settings somewhere to change core usage while trying to diagnose/fix my old CPU (which I think was dying). One solution was to limit core usage to the first six. That may have been it.

EDIT: SOLVED
run msconfig -> boot -> advanced options -> number of processors unchecked -> restart

I am very glad I booted up hwinfo on a whim, otherwise I would have been running 6 overclocked cores for no reason. Was also wondering why the system was generating so much heat when I had one of the best aios attached. Thanks again Dalai, sometimes it takes someone else to point out the obvious.
 

Attachments

  • Test Old System v2.jpg
    Test Old System v2.jpg
    511.9 KB · Views: 4
Last edited:
Top