B550 Tomahawk power reporting deviation (Ryzen 7 3700X)

[^B&]

I tested power reporting deviation during blender run, and minimum was 81.3%. This was at full stock settings


Then i ran the same test with -0.1V core voltage offset. Minimum value was 67.6%

Are these values bad? Can I do anything about it?
Why is PRD lower with negative voltage offset applied? Is measure with offset reliable?

 

[Martin]

Please read again the conditions for this test: https://www.hwinfo.com/forum/thread...er-reporting-deviation-metric-in-hwinfo.6456/
You need to be running at stock settings.

 

[^B&]

Please read again the conditions for this test: https://www.hwinfo.com/forum/thread...er-reporting-deviation-metric-in-hwinfo.6456/
You need to be running at stock settings.

Is Cinebench R23 fine too, or only Cinebench R20 is proper stessing software? How about Blender? It also maintains full and constant cpu load.
While Cinebench R23 run without minimum test duration, lowest PRD was 90%. Bios was set to (almost) defaults. Only thing I did in bios was RAM manual OC (not XMP). I guess this is fine result. Am I right?

 

[Zach]

I tested power reporting deviation during blender run, and minimum was 81.3%. This was at full stock settings
View attachment 5597

Then i ran the same test with -0.1V core voltage offset. Minimum value was 67.6%
View attachment 5598
Are these values bad? Can I do anything about it?
Why is PRD lower with negative voltage offset applied? Is measure with offset reliable?

Which one is giving you best results, and what is the core clock in each? Its clearly obvious that with offset in place you're getting better CPU temperature.
Sub 100% PRD means that your board is understates the CPU consumption to the CPU its self to push it more. But I wouldnt care too much, when the temp is below 70C.

82% PRD means that your CPU is consuming (CPU PPT) 107W and not 88W as the board reports but as I said I wouldn't care.
About the 69% PRD with the offset we cant really say anything because when you start to mess with the power management of the CPU the PRD value is completely useless. If you put down all the data you can understand it.
If we take the 69% for real then that means that the CPU is consuming 127W and not 88W that PPT reports. Tell me now, how it's posible to have better temperature with 127W against the 107W?

So... PRD means something ONLY with CPU on full auto/stock settings.

 

[^B&]

Which one is giving you best results, and what is the core clock in each? Its clearly obvious that with offset in place you're getting better CPU temperature.
Sub 100% PRD means that your board is understates the CPU consumption to the CPU its self to push it more. But I wouldnt care too much, when the temp is below 70C.

82% PRD means that your CPU is consuming (CPU PPT) 107W and not 88W as the board reports but as I said I wouldn't care.
About the 69% PRD with the offset we cant really say anything because when you start to mess with the power management of the CPU the PRD value is completely useless. If you put down all the data you can understand it.
If we take the 69% for real then that means that the CPU is consuming 127W and not 88W that PPT reports. Tell me now, how it's posible to have better temperature with 127W against the 107W?

So... PRD means something ONLY with CPU on full auto/stock settings.

CPU temperatures on stock settings are quite high tho. For example while Cinebench R23 thermal throttling test, it spikes up to even 82C, between tests. However with -0.1V offset it should stay below 70C or maybe a little above. My cooler is Be Quiet Dark Rock Pro 4. Other than that max peak temperature for Cinebench R20 run, on bios defaults, is around 70C.

Do you think I should keep -0.1V offset

 

[Zach]

I thought that was obvious from what I was saying. Yes keep it with the offset if you're not loosing performance, but dont mind what PRD is saying because the second you start taking CPU power stuff on hand, PRD has no value. The temp drop indicates lower power consumption and not higher as PRD is reporting.