Score:0

Idle power consumption increases after kernel 6.1.0.0.8 by 10-12Watts

in flag

I have noticed, that my computer, which I put together myself and I can give very specific details about it but I wanted to ask about the problem first. My computer in question is a Gigabyte Aorus Z690-Pro with an Intel Core i7-12700 (no letters), which I measured at around 29 Watts idle power consumption. So little later came the graphics card, a Sapphire Radeon RX 6600, so I could at least say, it can ray-trace and stuff. I put the timer to one minute before my screen goes black and the computer goes idle. At the desktop and nothing running, that computer uses 62 Watts to show a picture and some icons. But that's not the issue. My issue is that with all the Bios updates that I flashed in, the idle power consumption varied quite a bit on this one but until kernel 6.1.0-1008-oem, it was still withing the margin of what I would consider stable idle power consumption. With the card it went up slightly but I still count 32.1 to 35 Watts a very good idle power value for a computer that brings eight more cores than I ever had before and still requiring around 20-ish Watts less idle power than my former computer, featuring a quad-core with the Skylake architekture, named Kabylake. This one used 50-ish Watts with a Radeon RX 480. But now I have a gain of 10 to 15 Watts by just booting one of the kernel updates that folloewd 6.1.0-1008-oem. I was hoping one of the never ones would bring me back to 33 Watts idle power consumption but every single one of them causes an uptake of 10 to 15 Watts in idle power consumption. Please don't hit me, but I also watched my power consumption on Windoze or Reboot 11. With the latest developer preview Windoze update, whenever that Windoze goes idle, that idle power consumption goes down to around the same 34 - 37 Watts idle power. I have this measurement devise in my wall-plug, which shows me the power consumption of the entire computer down to single digests or 0.1 Watts. My question now would be, what am I doing wrong or thinking wrong or messing up some vital steps that I really cannot see. My tunnel-vision in that case only sees me booting up the next kernel just like every other security update I installed over the years.

Let me know if and how specific details you would need to answer that question.

EDIT: Add some turbostat information from the comments in a readable form:

Busy% Bzy_MHz IRQ   PkgTmp PkgWatt CorWatt GFXWatt RAMWatt
0.80  2580    23310 29     2.86    1.40    0.00    0.00
0.27   976    12290 30     1.65    0.19    0.00    0.00
0.26  1072    10882 29     1.58    0.13    0.00    0.00
0.20  1061     7175 29     1.55    0.10    0.00    0.00
0.19  1061     7109 29     1.54    0.09    0.00    0.00
0.22  1103     7428 29     1.55    0.09    0.00    0.00
0.18  1100     6875 28     1.54    0.09    0.00    0.00
0.19  1037     7065 27     1.53    0.08    0.00    0.00
0.19  1049     7383 28     1.55    0.10    0.00    0.00
0.32  1140    14631 29     1.62    0.17    0.00    0.00
0.36   972    15399 27     1.57    0.12    0.00    0.00


This is my current readouts, running kernel 5.15.0-73.80:
Busy% Bzy_MHz IRQ   PkgTmp PkgWatt CorWatt GFXWatt RAMWatt
6.65   838    86338 29     2.07    0.62    0.00    0.00
6.61   845    82833 30     2.13    0.67    0.00    0.00
6.80  1085    83161 28     3.87    2.42    0.00    0.00
6.53   861    77710 27     2.14    0.69    0.00    0.00
6.53   846    78460 28     2.14    0.68    0.00    0.00
6.56   852    78009 27     2.13    0.67    0.00    0.00
6.53   839    76298 28     2.05    0.60    0.00    0.00
6.54   833    76197 28     2.03    0.58    0.00    0.00
6.57   838    77727 27     2.03    0.57    0.00    0.00
6.53   838    76101 27     2.07    0.61    0.00    0.00
6.56   834    74944 27     2.04    0.59    0.00    0.00
6.65   896    79104 27     2.45    1.00    0.00    0.00
guiverc avatar
cn flag
Please refer https://askubuntu.com/help/on-topic, Ubuntu and *official* flavors of Ubuntu (https://ubuntu.com/download/flavours) are on-topic on this site. The on-topic link provides alternate SE sites for non-Ubuntu OSes. You don't mention anything specifically on-topic for this site (Ubuntu isn't mentioned), and don't word it as a "*problem to be solved*" (see https://askubuntu.com/help/dont-ask) Don't forget this is a Q&A site and not a Forum (*which answers your sort of questions generally better; questions are focused problems on a Q&A site*)
Doug Smythies avatar
gn flag
It would be useful if you could use turbostat (linux-tools-common package, I think) to report processor package power with the reference 6.1.0-1008-oem kernel and with the others. Suggest something like this: `sudo turbostat --quiet --Summary --show Busy%,Bzy_MHz,IRQ,PkgWatt,PkgTmp,RAMWatt,GFXWatt,CorWatt --interval 30` as a starting point. You would have to bisect the kernel in order to isolate the exact commit that introduced the issue, which I only know how to do for mainline kernels. I do not know what to do for oem kernels.
elita two avatar
in flag
I am sorry for the delay. Thank you for taking the time. I will do that and get back to you as soon as I have the results.
elita two avatar
in flag
Sorry in advance for a slightly longer response. For your earlier question, it is not Ubuntu, but it is based 100% of Ubuntu.
elita two avatar
in flag
Busy% Bzy_MHz IRQ PkgTmp PkgWatt CorWatt GFXWatt RAMWatt 0.80 2580 23310 29 2.86 1.40 0.00 0.00 0.27 976 12290 30 1.65 0.19 0.00 0.00 0.26 1072 10882 29 1.58 0.13 0.00 0.00 0.20 1061 7175 29 1.55 0.10 0.00 0.00 0.19 1061 7109 29 1.54 0.09 0.00 0.00 0.22 1103 7428 29 1.55 0.09 0.00 0.00 0.18 1100 6875 28 1.54 0.09 0.00 0.00 0.19 1037 7065 27 1.53 0.08 0.00 0.00 0.19 1049 7383 28 1.55 0.10 0.00 0.00 0.32 1140 14631 29 1.62 0.17 0.00 0.00 0.36 972 15399 27 1.57 0.12 0.00 0.00 ^C0.25 1047 3366 27 1.58 0.13 0.00 0.00
elita two avatar
in flag
This is my current readouts, running kernel 5.15.0-73.80:Busy% Bzy_MHz IRQ PkgTmp PkgWatt CorWatt GFXWatt RAMWatt 6.65 838 86338 29 2.07 0.62 0.00 0.00 6.61 845 82833 30 2.13 0.67 0.00 0.00 6.80 1085 83161 28 3.87 2.42 0.00 0.00 6.53 861 77710 27 2.14 0.69 0.00 0.00 6.53 846 78460 28 2.14 0.68 0.00 0.00 6.56 852 78009 27 2.13 0.67 0.00 0.00 6.53 839 76298 28 2.05 0.60 0.00 0.00 6.54 833 76197 28 2.03 0.58 0.00 0.00 6.57 838 77727 27 2.03 0.57 0.00 0.00 6.53 838 76101 27 2.07 0.61 0.00 0.00 6.56 834 74944 27 2.04 0.59 0.00 0.00 6.65 896 79104 27 2.45 1.00 0.00 0.00
elita two avatar
in flag
I am still looking for the kernel tools for kernel 6.1.1008, but those have been deprecated and deleted from anywhere I looked.
Doug Smythies avatar
gn flag
I am not understanding the reasons for your differences in the 2 turbostat outputs. Regardless they wouldn't explain the massive difference in mains power. Just as a test, I would try [the latest mainline kernel](https://kernel.ubuntu.com/~kernel-ppa/mainline/v6.4-rc5/). Since your OS isn't Ubuntu, this question will likely get closed. More for my own interest, what do you get for `grep NO_HZ_FULL /boot/con*`?
guiverc avatar
cn flag
Please refer https://askubuntu.com/help/on-topic, Ubuntu and *official* flavors of Ubuntu (https://ubuntu.com/download/flavours) are on-topic on this site. The on-topic link provides alternate SE sites for non-Ubuntu OSes. *Ubuntu based products are not on-topic, only Ubuntu systems*
elita two avatar
in flag
@DougSmythies, I get /boot/config-5.15.0-73-generic:# CONFIG_NO_HZ_FULL is not set /boot/config-6.1.0-1008-oem:# CONFIG_NO_HZ_FULL is not set /boot/config-6.1.0-1013-oem:# CONFIG_NO_HZ_FULL is not set /boot/config-6.4.0-060400rc5-generic:# CONFIG_NO_HZ_FULL is not set , when I do the grep
elita two avatar
in flag
I am sorry to report, the mainline kernel RC does behave the same as the other kernel releases after 6.1.0-1008-oem and my measurement device shows 43.2 Watts idle power consumption after the screen goes into standby or black. I already tested the mainline 6.3 but there was no change in power consumption either. I am at a loss.
Doug Smythies avatar
gn flag
I am at a loss also. Sounds similar to [this one](https://bugzilla.kernel.org/show_bug.cgi?id=217455). I think you will have to bisect the kernel to isolate the problem.
I sit in a Tesla and translated this thread with Ai:

mangohost

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.