Possible Alder Lake CPU power package bug?

papajo

New Member
So I did set the PL to 35 ok but still I find it strange that I got such a high score (especially considering the clocksreaching 5.2GHz) on cinebench while the average power was only 7 watts according to HWinfo

1653377190157.png
 
5.2 GHz was most likely reached for a very small fraction of time only. Check the average clock and effective clock values, those are much lower.
 
5.2 GHz was most likely reached for a very small fraction of time only. Check the average clock and effective clock values, those are much lower.
Still that score equals a ryzen 5800x (zen 3 16 threads) I mean ok I know that my CPU is faster but that much faster consuming only 7 watts? The r23 bench does not finish instantly it needs some seconds so its not that the CPU just mastered it in a split second and didnt have time to start "draining some juice" :P
 
Yes, that's very strange. Check which core were mostly utilized during the test, the E-cores only perhaps?
 
Yes, that's very strange. Check which core were mostly utilized during the test, the E-cores only perhaps?
All the cores were utilized, but again it wouldnt make sense simply because of the score, I mean lets consider what you asked me (only the 8 E-cores being utilized)
How is it possible for the E-cores only to be as fast as 16 threads of a 5800x? (it wasnt an exagerations that's how much a 5800x gets on cinebench)
And ok lets asssume that the E-Cores alone are equally as fast as the 16 threads of the ryzen, are they so fast while only consuming 7.5 watts?
If that was true it means that the 12900K a) is the most efficient CPU in the world even compared to an apple M1, b) that the 12900K is like 3 times faster or even more compared to a 5800x and again I know it is faster, but surely not that fast... I think the most plausible explanation is that there might be a bug with hwinfo reading the package power consumption of alderlake (maybe due to the e-cores being new tech? ) or at least being buggy if an alderlake has low custom set PL1/2 values
 
The E-cores are pretty effective/efficient but such a difference would be too much.
I'm too wondering whether the power reporting works properly in this case, do you have other means to check total power consumption like a wall meter?
I'm also curious whether the applied voltage offsets might be causing the issue with power reporting. Can you try to remove these offsets?
 
Back
Top