Created attachment 304284 [details] acpidump / lspci / dmesg logs I have a setup for measuring power consumption straight out from AC socket. I'm using following hardware in test setup: ASUS A88X-PLUS motherboard / Athlon X4 845 / 16GB DDR3 1600 a usb sata ssd attached as system disk, no VGA card Tested on Debian bullseye kernel 5.10 , bookworm kernel 6.1 and fresh stable 6.3.2 compiled by myself. Results are the same. All attached logs are from 6.3.2 After a fresh boot, without any interaction such setup consumes ~19W out from power socket. Doing a suspend to ram and after resume power consumption in exact idle use case is increased to 30W and do not come back to value before suspend. I was trying using powertop --auto-tune but it had almost no influence on the increased value.
Created attachment 304313 [details] Graph of mains power before and after suspend to ram After seeing an email on the linux-pm list about this bug report, and because I have a mains power meter attached to my linux test server, I tried this. I didn't notice any significant difference in the mains power or the processor package power before and after suspend to ram. I used this command: $ echo mem | sudo tee /sys/power/state Kernel was 6.4-rc3. Linux distro is Ubuntu 20.04.
(In reply to Doug Smythies from comment #1) > Created attachment 304313 [details] > Graph of mains power before and after suspend to ram > > After seeing an email on the linux-pm list about this bug report, and > because I have a mains power meter attached to my linux test server, I tried > this. I didn't notice any significant difference in the mains power or the > processor package power before and after suspend to ram. I used this command: > You have a whole different HW setup, Doug. (In reply to Alexey Kunitskiy from comment #0) > Your best chance of helping fix this is to perform regression testing: https://docs.kernel.org/admin-guide/bug-bisect.html It will take a lot of effort but no one but you can do it.