I literally ran the tests with hardware I have, if you see flaws in my testing setup feel free to point them out.
I would believe Adrian_Joachim here. Linux does not run a bunch of background tasks all the time like Windows. If he is using a properly tuned power profile, and really just has 2 terminals open in Sway (known for being really light weight and fast, although I would recommend wayfire personally) then I really don’t see why his power requirements in this state are not accurate.
Windows is abysmal in this case, but using a bunch of group policies and registry hacking you can get kind of closer but still not really.
What Lunar lake is offering in terms of power efficiency is really impressive. Everyone thought x86 could never come as close to ARM in that department. Happy to see Intel proving them wrong to some degree.
I would argue as well that if power efficiency is important to you, than maybe loosing the ability to supply your own RAM, etc. is an acceptable tradeoff. I have 4 framework mainboards that are based on ddr4 RAM. If I upgrade any of them to current generation I have to buy new RAM. This is functionally similar to some extent. You would be able to replace the mainboard though, and so while not as nuanced you would still be getting a lot of modularity and repairability. If Framework ONLY offers mainbaords with soldered RAM then that is a true step back (unless the industry mandates it). However, if you have a choice of socketed RAM but less power efficiency, or soldered RAM and more power efficiency (Intel claims that this saves approximately 40% more power alone) I would say Framework would be holding true to their ideals.
1 Like
The only reason I use sway is cause it took most of my existing i3 config and I am very lazy when it comes to changing something that works well enough.
Sufficiently deshittified windows should be able to get better numbers especially in video playback as it can still use the hw decoders and stuff more efficiently (it is improving on linux though, was a lot worse a while ago).
I can’t wait for lpcamm to make this tradeoff go pretty much entirely away, on package memory can still be slightly more efficient but it should reduce the difference enough to not really be an argument anymore. (Not that I would be mad if someone put a couple gb of hbm on a package in addition to lpcamm XD)
1 Like
I’ve gotten decent results in similar tests.
In similar testing to him except in Windows and with a FWL16 (his numbers indicate a ~61 Wh battery indicating a FWL13) I’ve seen around ~40% higher power in similar tests (desktop with 2 command prompt windows open, windowed YouTube playback, local high res video playback) but since the battery is also ~40% larger the battery life numbers are similar.
I wouldn’t be surprised if the reason for the higher power draw may be largely due to the larger screen.
Only things I’ve done are disabled OneDrive and some apps (that I installed separately) that wanted to run in the background when I didn’t want them to.
Windows has more bloat consuming power, however it also has more optimizations.
2 Likes
Pretty much, as I said pretty sure sufficiently neutered windows should outperform my numbers especially in video playback.
Totally agree. It is not popular because Windows is the popular thing to ignorantly bash on, (from a privacy and customer abuse standpoint it is totally warranted) but from a technical standpoint Windows is professionally developed and it is impressive.
That kind of performance (talking about Lunar Lake here) and power efficiency wont be seen on Linux for a long time from now. At least a year, and even then it, most likely, will still not be as efficient. (don’t worry though the Linux crowd will have an ample supply of excuses for why this is the case.)
1 Like