12th gen battery questions on linux

auto-cpufreq, TLP, etc

I’m interested in seeing results on battery life in Linux with a 12th gen chip. specifically with use of tuning like auto-cpufreq, I’m wondering if this program or others like it can cope with the new hybrid architecture of 12th gen.

Disabling Cores

also curious about results after disabling 2 of the P cores which should essentially turn the CPU into a U series processor (which is what I wish framework went for in the first place :stuck_out_tongue: although if disabling cores saves battery then I won’t mind).

Disabling Turbo Boost

I’m also interested to know how much power is saved by disabling turbo boost and how much this effects usability for day to day tasks. I feel like these CPUs are already so fast compared to the i5-7200U in my current laptop that even a severely crippled chip could still feel fast although I suppose turbo boost is all about being snappy.

IO / Expansion Cards

I’m curious about difference in power draw between all of the IO modules, specifically USB-A and HDMI. I’ve heard conflicting information about the USB-A power draw. I’m wondering if I should order an extra USB-C to use most of the time instead of the HDMI if it adds a little extra battery life.

SK Hynix Gold P31 SSD

this is definitely becoming too much to ask but also if you have a sk hynix gold p31 ssd that result would be more relevant.

i5-1240P vs i7-1260P

also if anyone has insight into actual numbers for i5-1240P vs i7-1260P for battery life that would be appreciated. I understand that the latter should be better but concrete results would be awesome.

Linux Kernel Version (5.17 or 5.18)

another question of mine is how much is kernel version going to change the battery life? it seems that newer versions have much more fine tuned scheduling for 12th gen with the Intel Hardware Feedback Interface (HFI, I think in windows it’s called the thread director).

11th Gen Questions

if there are only 11th gen customers here I’m still curious about how much battery life you get using auto-cpufreq and maybe TLP with the SK Hynix Gold P31

Thanks!

I don’t really expect answers but thanks anyway!

4 Likes

USB-C cards are passive and consequently energy draw is zero.

4 Likes

I think he was asking about USB-A specifically, which is a question I had as well. Early on, I saw reports that both USB-C and USB-A expansion cards were unpowered pass-throughs, but people now seem to treat USB-A as another power draw. Has someone confirmed that? In the early days of USB-C, I was under the impression that going from A to C was a passive adapter, but that doesn’t mean there aren’t some active forms out there too.

1 Like

I believe the consensus on the forum is that USB-A modules are active and do draw power, even when they probably shouldn’t (i.e connected but not in use).

2 Likes

that’s a shame, hope that is fixable; perhaps a v2 USB-A port might be in order if it can’t be fixed in software. might end up getting 4 USB-C now! personally I feel like the customizable ports are for the most part a gimmick, or at least the priority should have been greater battery life. the tradeoff of more used space and power being needed for many of the modules doesn’t seem worth it to me. but I’m biased because I’m one of the people who would want a pretty standard port selection, I don’t mind a dongle for something like ethernet.

1 Like

The Linux kernel has seen an improvement in batterylife for alder lake with 5.19 as well (for the cost of performance though)

For reference, these tests were made with 5.18 (iirc):

It is not a framework, but benchmarks for 12th gen are pretty scarce for now…

2 Likes

I’ll take that! I think CPU performance is massively overhyped for day to day usage (don’t get me wrong, I could benefit from faster CPU as I compile rust btw. but the difference won’t be massive) but battery life is still such a big bottleneck on mobile devices.

1 Like

The usb-a module is just a pass-through, no active components.
There are a few caps and a resistor, none of which draw anything.


6 Likes

Ah! This is the kind of info I was looking for. Yes, from those photos I would be amazed if USB-A modules were hurting battery life.

1 Like

Very interesting. I think the person on this thread are refering to the “1,5Watt/hour” that has been measured using modules.
Now it seems that USB A & C are not the culpride.
This would let us just unplug the HDMI when not in use… maybe we can deactivate it from software in linux then.

Thks SImon,
That s very good news, ~ 5% less speed (20% in IA) , but 20% less power drawned for the same taks. That s exactelly what I was expecting from “efficient” cores.
That s really cool, especially thinking those are crunching number benchmarks. WHere the CPU actually as very little time to get idle!
This possibly means that the battery life could be extanded more than 20% CPU wise. Looking forward to a “battery test” with this new kernel (browsing the web and videos).
Those 20% would mean we could get back the 1H battery life, we had with 11th gen, while having better performances.

1 Like

Awesome! Do we know if the DisplayPort extension is passive or not?

Yeah this is very promising. In general is it considered a bad idea to upgrade to a new kernel without support from the distribution? This seems like such an uplift that waiting for the distro to update to it seems painful. I think fedora 37 which isn’t even out yet will be based on kernel 5.18, I’m not sure how quickly Pop!_OS tends to update its kernel. I suppose this is a situation where running Arch pays dividends.

Upgrading kernels is totally fine and a great thing to learn how to do. I’m sure there’s a PPA which has the latest kernels but it may be worth learning to compile your own. You can always have multiple kernels as long as your EFI/boot partition is big enough so you can switch back and forth (I like to have a mainline kernel available). The Arch and Gentoo wikis are a decent reference for boot loader and kernel stuff.

1 Like

That’s good to know, thanks!

Depends on what you are doing with your system. It going to be my main rigg. So I am not planning on testing kernels. Anyhow you will be able to boot with one or the other kernel via GRUB. SO you can always fall back to a working kernel.
I personnaly would better go for next Ubuntu 22.10 in October. Even thought this would put me in the non LTS wagon until 2024 !

Found also this for evaluation , just copy the iso on a USB key and off you go to test the new kernel :smiley: https://cdimage.ubuntu.com/daily-canary/current/

I am on Mate currentely, “for a restrospective future” !!! <3 , the daily is there: Ubuntu MATE 24.04 (Noble Numbat) Daily Build

1 Like

Thanks
I probably won’t run Ubuntu on my framework, I don’t like the direction it’s headed with snaps and I think it’s not as interesting as what System76 are doing with Pop!_OS or what you can get with more up to date stock Gnome on Fedora. I’m saying this as an Ubuntu 20.04 user on my current laptop.

Plugging in one USB A module on the 11th gen board.


This is the drain adding a type A and then HDMI.

4 Likes

This is so interesting, have any people on the framework team addressed this as far as you know?

It not so much the USB A but clearly it has to be recognised as being not the deafult USB C so it will ‘cause’ if not ‘draw’ power.

However why that is over 1W rather than 0.001W is the issue as far as I can see.

1 Like

Not that I am aware. Quite a lot of folk have noticed anything not type C drains additional power at least on 11th gen.

1 Like