Yes, this is using the Balanced profile, the Power Saver profile is just a little better, but it doesn’t make sense due to the performance loss.
I don’t even know anymore, can I try and use a different CPU scheduler? Or is a lot of that embedded in the CPU now? I’ll also try downgrading the kernel, maybe there’s a problem there.
Wot? how did you measure that? (please don’t say powertop)
Easy way to test, uplug the display and see what the draw is then (via ssh or something).
What I would do to measure minimum consumption is (on a fresh boot with no applications running) go to desktop, enter battery-saver mode, turn brightness to 0% (5% works too if you’re more comfortable with that, it’s not actually a huge difference in draw, but a big difference in viewability), open a battery monitoring software (e.g. powertop or btop) and then check on it in 5 minutes after it’s settled.
If I were in your shoes and found that power draw is still high after that, I’d be using htop to check if there are any processes using more than 1% of the CPU, and use nvtop to see if there are any applications using the dGPU (or better yet use cat to see if the dGPU is in d3cold with cat /sys/class/drm/card1/device/power_state).
Edit: There some more info in this thread: Minimum Power Draw Competition - Framework Laptop 16 - Framework Community
You can also try powertop --auto-tune
I believe what I had done was got the average power draw during 100% brightness, 0% brightness, the power draw after issuing some “display off” command I found which turned the screen black, and the power draw when utilizing an external monitor. Powertop may have been used to read the wattage value, but no I was not using the practically useless draw estimates it reports after calibration.
I may have also looked at a measurement after the display timed out. I know I reported the 3W figure somewhere, but I don’t know if I recorded the method of measurement elsewhere on the forum. I do know that the difference between 60hz and 165hz is ~1W
@Adrian_Joachim I found where I recorded this number, and it appears that 3W may have actually been for my old Razer Blade 15… but I distinctly recall measuring this on the FW16. It’s possible I’ve been mistaken, FYI.
Even more reason to test with the display unplugged XD.
But yeah on the min idle front progress is slow, my fancy new fw 13 can barely beat my almost vintage t480s in min idle (and after a lot of fixes the framework can at least match it in video decode power XD). When the cpu is actually doing something it’s a bit of a different story.
When I started my battery life benchmarking I was comparing against my T470, which idled at a phenomenal 2.7W, but I recently realized I may have been misinterpreting that answer. My T470 has 2 batteries. I suspect that the thinkpad actually pulls from both batteries at the same time, but only reports the wattage from one of them. A clue that supports this theory is that I can’t remove the “external” battery and replace it with a fresh one live, the device shuts off if I do that. I think the Thinkpad may actually be drawing double the power that’s reported by the battery usage.
Yeah, but the idea of the min power draw test is to determine the floor/max possible battery life. That way, any deviation from that path is deliberately chosen by you as the user as a way of saying “I wish to exchange this energy for some value” as opposed to it just being sapped away by some arbitrary set of conditions and default configurations.
If it’s the same power bridge tech as in the x270 (or 240 or 250 or 260) then no it drains one then the other not both at the same time.
That may be a defect, being able to live swap batteries is like the whole point of the power bridge thing.
Measuring min idle definitely has a lot of value (which is why I am doing it XD).
In the case of the t480s vs the framework 13 they are basically equal in min idle and video playback battery life (with the fw 13 coming out slightly ahead because of the bigger battery) but the fw whipes the floor with it as soon as you actually do something somewhat compute intensive.
battery life is excellent on my fw16/win11/no dgpu, i get around 6 hours of battery life, 7 if its all light usage
I have found things that affect battery life are:
You could probably save more power if the FW16 keyboard supported power save modes when not pressing keys, but there are bugs there. It might also be due to the usb api is not a good fit for keyboards and various power save modes.