I do not have the dGPU, although I have seen reports elsewhere that the idle power consumption is basically null when in d3cold so I’m not currently inclined to blame the dGPU further.
Out of curiosity, could you try running powertop --auto-tune, rerun the idle test, and see if that made a difference? Note that depending on what version of packages you have, running that auto-tune may make the FW treat the keyboard like an external and so it may need to be woken up after idling when using powertop this way.
I actually have not used powertop at all so far. I’ll do that now. I did some further testing up to this point and managed to reduce power consumption to around 8W by removing the numpad module. I am not really going to live without my trusty numpad, but it is still interesting to know that it has a measurable impact on power consumption.
That’s quite interesting. I have the numpad as well but I haven’t tested the power consumption of it specifically. Maybe if I get the time this weekend I’ll pull the mainboard out and use my usb-c power-meter to measure the draw of each component without relying on the software measurement.
Ok. With all expansion cards installed (including numpad) my system now idles at around 5.8 to 6W according to powertop. Considering I started at 14W idle, thats a pretty nice improvement. I will test once more without my modules and see how far I can push this, but I am honestly quite happy with that.
Update 1: I only wished I could monitor Temps in the task bar without having this prevent the dgpu from sleeping.
Update 2: Ok, now, removing the numpad actually has 0 effect on idle consumption. I guess that is because it now goes to sleep properly when not being used, unlike before.
I have a minimum of 7W with only GNOME Console opened.
7940HS, 64GB of RAM.
Minimal brightness, 0 load average, power-saver in powerprofilesct, disabled wifil.
The real life scenario sounds about right to me as well. With a browser open and normal screen brightness (20 to 30%) I am usually sitting around 10 - 12 Watts.
This definitely isn’t at idle (browsing this post, listening to music on Spotify with Bluetooth etc.), but I’m wondering why my keyboard is using so much power.
Normally, my laptop idles around 10W, though I haven’t tried turning all the knobs like reducing brightness to 0%. Will do that this evening to get a proper baseline.
I guess the high keyboard power usage is partially caused by me disabling idle. Otherwise, every first keypress after a few seconds gets ignored or sticks until pressed again. Though I use my keyboard a lot and it’d be nice to have it use less power during usage, too. It’s the blank layout, therefore I always keep the backlight disabled.
tick_nohz_handler is also often one of the top consumers - from what I’ve gathered online, it’s just some kind of kernel event loop and therefore fine. But I still wonder if it’s normal or still a bit excessive.
For reference, I’m running Arch with Hyprland, the Ryzen 9 and dGPU with 64GB@5600MHz RAM, currently in power-saver mode. After disabling wayland compatibility in electron apps, the dGPU finally suspends properly. Therefore, it shouldn’t be much of an issue. My goal is to make the battery last long enough for a day of university (I do have a power bank, but the longer, the better).
I’m not sure how much I would trust the power estimates of PowerTop, I’ve not found them terribly useful myself even after running the calibration. If you’re running RGB/White backlighting on the keyboard though, that will absolutely draw more power. I keep mine off unless it’s night or I’m plugged in. Unfortunately I don’t remember how much the white backlight draws, but I probably recorded it somewhere on these forums.
What I would do in this scenario is
make sure that the 18W is a stable load, e.g. stays at ~18W for a minute or so
Look at cpu usage to identify any high usage processes. E.g. I’ve personally diabled baloo because it ran far too often and far to intensely for my liking and I didn’t seem to be getting any benefits from it.
I would remove usages one by one to estimate how much each one is costing me. E.g. I’d close spotify and turn off bluetooth and see how much it drops, then I’d close the browser, etc.
If I had to guess, spotify is what’s killing your battery life. That’s been my experience with it anyways.
Like I said, no backlight on the keyboard (no point to it, it’s the blank one - therefore no RGB, either).
Spotify definitely draws a lot, as do other electron apps. But even with everything closed, I still can’t drop below 10W. Been an exhausting day, so I’ll postpone further testing to tomorrow.
Here’s my simplified overview, assuming you’re on linux with up-to-date packages (e.g. via Arch)
I have no dGPU. Some uses have issue with the dGPU turning on when it’s not needed (i.e. not in d3cold state), e.g. during web browsing. You can run the following command to see what state the dGPU is in at idle without waking it up: cat /sys/class/drm/card1/device/power_state
I use power-profiles-daemon to enter a battery saver mode. This is where the majority of savings come from
I use powertop to set power saving parameters on a number of other pieces of hardware. Note that this does have some downsides, e.g. setting the keyboard into sleep mode means it needs to be woken up before typing (so I don’t use it on that device).
I know (or at least knew and forgot) every automated process that runs on my machine. Running Arch linux means I’ve had a lot more influence over exactly what is happening and when, so I have confidence there are minimal background processes running that I don’t need/want.
Finally, my power usage figures come from light usage. A text editor and a web browser are often the only things running on my machine. Heavier load, video content, or etc. will naturally increase battery draw. E.g. playing Stardew Valley draws about 13.5W to 14.5W on my machine.
I set up a bunch of this stuff and was able to hit 13.1 watts at idle without changing my screen brightness. Not quite as impressive but still plenty of battery life for what I need. Especially when compared to the 2-3 hours I got on windows.
Now that GNOME supports variable refresh rate displays, I tried it out today and it lowered the power usage by a bit (I don’t remember how much exactly, sorry) without having to bump the screen down to 60hz. Variable refresh rate - ArchWiki
