I’m using gnome-terminal. I tried checking powertop over SSH with all apps closed and I get the same reading as with gnome-terminal running.
I’ve set display refresh to 60Hz and turned screen brightness down to the lowest setting. With no apps open, all ports removed & BT disabled I’m now seeing 6 watts consumption on battery. Better but on my 8 year old Dell XPS15 I get 5 watts with the brightness @medium when using Chrome.
Are there other configuration settings that can shed another ~2 watts?
That’s a bit apples and oranges and difficult to equate F16 has a higher resolution display than existed 8 years ago, you might have more RAM, different SSD technology back then etc.
You can experiment with some of the other things in powertop under the tunables menu. Some people in the F16 forum have found some success getting lower. You should probably look for that post in that section and comment there.
What is better now in 2024, tlp or ppd?
tuned + tuned-ppd works the best for me. 4-5W browsing
I don’t know what’s going on I’m using Fedora 39 with ppd and the turnables have lots of Bads, and tune them doesn’t work, they will go back to Bad upon reboot, even use sudo nano to text edit those “on” to “auto” won’t work
Please refer to the following page for applying powertop tuning parameters permanently: Powertop - ArchWiki
I have yet to see any evidence that changing the remaining things that are marked bad by powertop make any measurable difference when toggled.
If you think they’re actually helping, can you please collect data from a predictable workload with no change followed by with a change. In order to make a case to change a default in upstream software you’ll need to be meticulous and only change one knob at a time.
I just measured. When expansion cards except usba/usbc are removed, the idle power consumption is about 2.8W at lowest brightness regardless of powertop.service is enabled or not. When HDMI/microSD cards are installed, the idle power consumption is about 4.5W w/o powertop.service, enabling powertop.service reduces about 0.5~0.6W however it makes the wireless USB mouse laggy, touchpad is not affected though.
EDIT: inconsistent results, need more tests
Can you find the same improvement with just the SD card added and then turning on Auto suspend for JUST THAT CARD? I would think so; the HDMI card is already supposed to be put into auto suspend from a systemd quirk.
If the microsd card auto suspend improves things can you please check if it still functions properly? Most notably; if you put a card in does it automatically mount?
Unfortunately I can’t get a consistent result, after disabling powertop.service and rebooting the computer(got a bunch of Bads on powertop again), the discharge rate is still 2.83W. I tried rebooting to another OS then back (dual boot) unplug HDMI&SD, reboot, then plug them back, nope, still 2.83W not 4.5W. I have no idea what has changed persistently. I think the previous 3~4W idle is due to Fedora not resuming lowest brightness after rebootting, had to lower the brightness again to get consistent result
Yes, regardless of powertop.service.
I can’t remember when and how it was done, could you explain how?
How is there not one yet? I was very surprised I had to scroll through so much of the forum to find it hidden away… Like this is a secret only reserved for the chosen few😂
Bless us normies with this knowledge🙇
Scroll to installation guides.
I did notice that your ppa was in the guide today after I just installed it and it seemed familiar.
IDK if installing/uninstalling cpufreq messed it up or why my adaptive backlight management never worked with that (maybe I never actually installed it and only installed the OEM kernel from the guide…)
But now it seems to be working and for video playback, I’m getting 6.5w on 43% brightness instead of 6.5w on 10% brightness.
Are those good numbers ? Because I don’t think everything is in that guide, I noticed you talking about powertop auto managment or something and other things that I didn’t find in there…
For the matter of suspend, I can confirm that suspend has been working fine on 6.8 kernels as well. Currently on 6.8.2-300.fc40.x86_64 myself.
With a web browser and code editor open and nothing compiling, it looks like I’m sitting pretty much at 10w according to powertop, with --auto-tune already run previously. Any suggestions for reducing? I’m on Fedora Gnome.
Can you link to more reading on the fractional scaling power consumption? I just did a test on my own system and could not reproduce any meaningful difference in consumption between 1.0 and 1.5 scaling in Sway on Fedora 39.
Is this compositor independent? I.e. does it matter if I use Plasma, Gnome or Sway?
Yes, gnome uses mutter, kde uses plasma etc.