I recently added the dGPU to my Framework 16 and have noticed something I’m having a hard time pinning down. Without the dGPU, my FW 16 would reliably draw ~7W with spikes to 10W, netting me 12-15hrs of comfortable battery life. Ever since I added the dGPU my battery life has been closer to ~6hrs, but when I check on the dGPU it’s always in d3Cold state but my power draw is averaging 12W, with dips to 8W if I let it idle enough. My impression is that this is not a power draw delta the dGPU should be capable of when in the d3Cold state, and thus I’m obliged to consider that there is something else going on that’s drawing power, but I can’t shake the feeling that since the only drastic change that’s happened is the addition of the dGPU that it’s somehow related to that.
Has anyone else had a similar experience and perhaps have any insights how I can recover my prior battery life with the dGPU installed and idle? My data and prior knowledge does not seem to be in agreement with my experience, so I’m hoping someone has my missing puzzle piece.
One potential explanation is that applications are waking the dGPU momentarily when opening or switching tabs or something and it’s happening frequently enough to make a difference.
Could you monitor the state for a while and see if that’s the case? Maybe you can narrow down what app is doing it if so.
I was using watch to track the gpu’s state for about an hour while I was doing some non-gpu work and it stayed (to my surprise) in d3cold the entire time, despite me seeing the higher draw. I am working on the assumption that 1s would be a frequent enough interval, but even if I use a waking tool like nvtop, it reports the dGPU is only drawing 1W.
Without the dGPU, my FW 16 would reliably draw ~7W with spikes to 10W, netting me 12-15hrs of comfortable battery life.
Wait, what? I get 13W when idle (measured just now with powertop right after booting into a tty) at 45% display brightness at powersave mode, and about 17w when I’m doing something moderately light.
I think I’ve partially figured out what’s going on. It looks like both my web browsers (Brave, LibreWolf) are now consuming much more power than they were before. I think I’ve seen a related post somewhere that I’ll have to hunt down, I’m not sure if this increased power draw is due to a software change or the hardware change I made, but in either case I know the dGPU is not being used for the browser (according to nvtop).
My goal with this thread was to gather information on if anyone else was seeing similar issues, thinking. I am now thinking that this may be a general power matter for linux devices, and the thus this thread is not necessary anymore.
Update 2025-01-27: I took the dGPU out and ran some simple tests, saw the same battery draw, so for me that confirms my issue has nothing to do with the dGPU.
Since firmware 305 I have the same issue. I do not have a dGpu. But I have installed the M.2 adapter. 303 had better battery performance. But I did no tests to proof that.
Now that I think about it, I think I did the dGPU upgrade the same time I did the BIOS update. My gut tells me that’s not the issue, but I have nothing more to back it up one way or the other than that.
I’ve removed the reference to the GPU from the thread header, as whatever I’m experiencing does not appear to be related.
The only thing I am certain of, is that my battery isn’t lasting as long as it was. It’s very noticable going from 15hrs to 6hrs.
I found a number of places where L1.2 was mentioned in my output of that command (32 to be exact), but I don’t know if it’s saying it’s just supported or also enabled. I’ve attached the full output to this pastebin
As far as I’m able to tell, power draw behavior stays the same whether the dGPU is installed or not. Looking at output from nvtop and powerstat, it would appear that both cpu and gpu are pulling more power than they used to when under light to moderate loads.
I’m not sure how much those power measurements can be trusted, but I’m seeing at least 20 extra watts of draw under some conditions
Yeah I think I am gonna try doing that when I’ve got the time to sit down and run through it. Hopefully it is a kernel related issue, because then I’d at least have a work around.
What DE/WM do you use? I noticed my cpu usage was pretty high (5%) for my workload (which is normally less than 1%) and found that a KDE Plasma thread was using 26% of a core, which is quite a lot. On a hunch, I exited my desktop session and switched to a plan TTY shell, no DE running and found that my power usage was closer to my prior usage, it immediately dropped to 9W draw, which is a little higher than the 7W I’m used to, but that could be passive pull from the dGPU or something. This is the first clue I’ve gotten that could explain a power increase.
Curiously, I noticed my nvtop report also looks strange, considering the only thing open is Librewolf with the tab used to create this post, and my text editor. Usage seems pretty high.
Update, I just tested running gnome (wayland), and my power draw immediately dropped down to expected levels.
Update 2: lxqt exhibits the same power behavior when using kwin/mutter window managers. On kwin, battery draw is high, on mutter it is low. I also tested open-box and it appeared to be somewhere between the two.
Through experimentation with a few different DE’s/WM’s, I have become confident that the power draw difference is somehow related to the window manager in use. Kwin is pulling an insane amount of power, regardless of what DE is loading it, and every other window manager is pulling significantly less power. Initial testing suggests that there is a decent spread of power-draw amount across the different window managers.
).
