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

confused. will read more and test. i have the lowest power consumption ( measured from wall with meter) with a fixed all core oc (4.2 ghz) + 0.05 vcore offest in bios (1.1v) pbo, and all other boost DISABLED, rest all auto, using the AUTO settings? increased TEMPS at load + lowerboost and more power consu . ???


edit. clocks / temp idle
 
Last edited:
You follow now power deviator? :)

agentnathan009 said:
Sorry, I was wrong. Vacuum tubes have insulation effect so heating overcomes that insulation effect allowing electrons to flow.

Per AnandTech article about this subject the electrical leakage has been mitigated quite a bit so it isn’t as bad as it was years ago.

Nope, for Zen 2 it is the same max temp.
Click to expand...
Good to know, where did you find that AMD info? [edit] Nm, saw that it is on the page I linked after all! lol

Re leakage, ah ok, interesting, I wonder how much of an effect it has now then?

**********************************************

Finally tested my main rig at stock settings, so my MSI B450 Gaming Pro Carbon AC (stupid long names!) on stock settings bar memory settings, running CinebenchR20, I get a PRD of ~99% :) (score 3499).

Just for reference, when I had limited the CPU PPT to 65w (default is auto, max 88w), and to show not all max loads are equal.
PRD :-
Rosetta@home ~106%
LHC@home ~98%
CinebenchR20 ~102% (score 3327)
 
Last edited:
Hi,

My Asus Crosshair VII Hero Wifi (running latest bios 3004) with a 3950x is reporting ~75% of Power Reporting Deviation under full load (Cinebench R20), but I also undervolted my 3950x with a negative offset of ~0.825V
(undervolt optimized checking with cinebench runs and assuring to still have full performance on single and multithreaded).

Would undervolting not have impact on this Power Reporting Deviation? As under full load, the CPU does now indeed pull less power due to the undervolting ???
 
Err, it would have an effect, and yea that might be why you only have 75%, I'm no expect on this though ;).
Anyway, the intention is for it to be tested at default settings.
One way or the other, over/underclocking or over/undervolting will override the default power usage.
 
Could some interpret my results? I'm lost a little bit.

I have 3700X on MSI X570 Gaming Edge WiFi, overclocked manually. CPU Ratio is 42,00 (4200 MHz) with override volting 1,3875 V (it shows 1,38V-1,4V)

During Cinebench R20 Power Reporting Deviation is 160%. Is it good? bad? should I be worried?

What it means, when is a lot over 100% during test?
Untitled-1.jpg
 
mephrite said:
Could some interpret my results? I'm lost a little bit.

I have 3700X on MSI X570 Gaming Edge WiFi, overclocked manually. CPU Ratio is 42,00 (4200 MHz) with override volting 1,3875 V (it shows 1,38V-1,4V)

During Cinebench R20 Power Reporting Deviation is 160%. Is it good? bad? should I be worried?

What it means, when is a lot over 100% during test?
View attachment 4685
Click to expand...
That you motherboard reports pulling much more power than it normally would in a full load case...

Honestly, your overclock is pushing way too much voltage in an all core load for a 3700x on 7nm...
A better way to overclock is to actually undervolt the CPU with a negative offset and then let PBO take care of the rest...

I use my 3950x with a low load line calibration (allowing for higher voltages when idle/low loads in single core usage & lower voltages in all core loads --> lower power, thus higher boosting capability till it reaches thermal thresholds) and with a negative core voltage offset of 0.825V...

Every step of negative core voltage offset needs to be validated with a few cinebench runs to assure the performance does not drop when going to low in voltage. As with Ryzen 3000 the PC does not really crash, but the CPU downclocks on microscale level which cannot be seen by normal frequency monitoring tools other than just benchmarking the performance...
 
Well, my overclocking voltage was lower than default setting in bios. Don't know why, but my motherboard on default sets 1,46 v (in game boost mode & normal), so I lowered it to 1,38V.

Now I lowered to 1,35V and during Cinebench test Power Reporting Deviation was only 146% now... But also I had worse Cinebench result by 70 points. I will try with 1,3 V....
 
mephrite said:
Well, my overclocking voltage was lower than default setting in bios. Don't know why, but my motherboard on default sets 1,46 v (in game boost mode & normal), so I lowered it to 1,38V.

Now I lowered to 1,35V and during Cinebench test Power Reporting Deviation was only 146% now... But also I had worse Cinebench result by 70 points. I will try with 1,3 V....
Click to expand...

Key is to not manually set the core speed, but let PBO do it's work... Next to that, I would suggest to not use a fixed voltage, but a negative offset voltage, that still lets the motherboard control the voltage, but you just lower it with the offset...

Then if you found the sweetspot you can start playing with the PBO limits and then let the motherboard boost higher if you want. (I did not have the need with my 3950x)...
 
Infernoken said:
Is this screenshot taken at full load with a tool like Cinebench R15 or R20??? It states only 40W powerdraw while your max is 60W?
Click to expand...
Nope it's on idle. Should I do it during a stress test?

Edit: ok just read it, sorry. I've uploaded also the correct one.
 

Attachments

  • Screenshot (81).png
    Screenshot (81).png
    54.6 KB · Views: 19
  • Screenshot (82).png
    Screenshot (82).png
    1.6 MB · Views: 18
Last edited:
Martin said:
Oh dear, how many times do I have to repeat the same over and over again...
Click to expand...

Hi Martin, any view on my question I also asked on the Power Reporting Deviation figure while undervolting?
Can this affect the Power Reporting Deviation ???

My previous post on this:
Infernoken said:
Hi,

My Asus Crosshair VII Hero Wifi (running latest bios 3004) with a 3950x is reporting ~75% of Power Reporting Deviation under full load (Cinebench R20), but I also undervolted my 3950x with a negative offset of ~0.825V
(undervolt optimized checking with cinebench runs and assuring to still have full performance on single and multithreaded).

Would undervolting not have impact on this Power Reporting Deviation? As under full load, the CPU does now indeed pull less power due to the undervolting ???
Click to expand...
 
According to the latest beta (v 6.27-4190), the Power Reporting Deviation for my Asus ROG Crosshair VIII Hero (X570) with my AMD Ryzen 7 3700x is 82.4% under full multi-core load in Cinebench R20. I'm on the latest BIOS (v 1302).
 
marcotini said:
Ok lol funny that the result is the same, both in idle and under stress. I've uploaded the screenshots.
Click to expand...

LOL... Screenshot is again not valid... As this is taken after the test and not during...

Actual figures should represent the system under load, so during the test and not just after it...
 
During was always the same, look at average value. Here during:
 

Attachments

  • Screenshot (85).png
    Screenshot (85).png
    730 KB · Views: 17
Last edited:
Infernoken said:
Key is to not manually set the core speed, but let PBO do it's work... Next to that, I would suggest to not use a fixed voltage, but a negative offset voltage, that still lets the motherboard control the voltage, but you just lower it with the offset...

Then if you found the sweetspot you can start playing with the PBO limits and then let the motherboard boost higher if you want. (I did not have the need with my 3950x)...
Click to expand...
OK, I set negative offset voltage 0,0125 V (lowest I can get) with PBO enabled - now Power Reporting Deviation is 85% during Cinebench. :D

Unfortunately with PBO my score in Cinebench is like ~100 points lower...
 
You must log in or register to reply here.

Similar threads

M
Pretty significant difference in power readings (Package Power vs VR Power) ?
Replies
3
Views
48
Martin
Martin
Z
Incorrect GPU power consumption
Replies
4
Views
132
Martin
Martin
K
Can someone explain like I am 5
Replies
10
Views
686
keblin
K
R
Getting unusual readings on the HWinfo
Replies
9
Views
112
rtochallenger
R
T
power reporting deviation %1100
Replies
5
Views
324
They not like us Wop wop
T
Back
Top