For me with s2idle (and nvme.noacpi=1) laptop lasts only ~16-18 hours in sleep, whereas with deep it lasts 2-3days; I’ve also noticed that in s2idle the laptop keeps warm, whereas with deep it cools down to room temperature (i.e. that there seems to be a difference in what systems get suspended too); In any case… It’s extremely unlikely that I will ever buy or recommend framework to anyone again, even if you completely ignore all the documented sleep issues they have, purely on the basis of poor customer support and non-existent communication to their customers of significant changes. Also, even with 3.20 their firmware still reports “deep” as a supported state, however if you enable that, it soft-bricks your laptop requiring a forced reset by taking the battery out. This is not acceptable.
Hi,
I think I understand now why framework has been silent on the s2idle issue with the newest BIOS. They were working on a different solution already.
Look at these 2 posts on mastodon:
TLDR: They are working on a tool to automatically enable hibernation and be independent of suspend altogether. The key information is this part
Now, that said, hibernate would address battery drain while in suspend for Fedora (and later Ubuntu) as we’re not relying on suspend.
We’re putting the lift on hibernate instead of only relying on suspend.
While I think this new tool is a fantastic thing, it kind of throws the people not using Fedora/Ubuntu running Gnome under the bus.
I assume that it will all be open source, so there would be nothing preventing anyone from porting it to a different distro.
True, but if I understand it correctly it is essentially a GNOME Extension right now and it requires BTRFS. So it only makes sense for new installs. So it is in its current form not a generic solution for all Linux systems/distributions. Others will have to fiddle and do it by hand.
And suspend is still broken for my system…
Well, they only officially support Fedora and Ubuntu, so that’s nothing new. And if it requires a particular partition setup then the logical time to do it is on initial install.
Setting up hibernate is not that difficult. It takes a bit of effort, but it’s worth it. I have been very happy with suspend-then-hibernate on my systems.
Maybe we should take it to a different thread, but what is broken with suspend on your system, and what distro are you running?
I am in the same club as these fine people over here:
As this thread is focused on BIOS 3.20, if you have issues with suspend and would like help with that, please move to another thread or open a new one.
Well, my issue with suspend was introduced after updating my from BIOS 3.17 to 3.20. Which soft-bricked my computer until I switched suspend mode from “deep” to “s2idle”
Deep suspend was working perfectly fine before and the battery drain was tolerable. Now it is not anymore as most of the time the laptop doesn’t really go to sleep - it just pretends it does by flashing the powerbutton. But if you listen closely you can still hear the fan spinning. Which tells me it is not sleeping at all. In that state it lasts maybe a few of hours instead of days.
Using hibernation is fine by me - but it isn’t fixing the underlying problem. It is just covering things up.
But fine, I will watch how this goes and if framework is providing a proper solution for this with the next BIOS version and then I decide if it is worth spending more money.
@lvdd deep sleep won’t be fixed anymore.
I was in contact with support and since 3.20 deep sleep is not supported anymore:
“Spoke with engineering for clarification on deep sleep not working. Yes, you will be using s2idle as we have found that the later BIOS are not supporting deep.”
That’s a pity for everyone with this issue, especially since there was no mention anywhere. At the least they added it now to the “Known Issues” for bios 3.20.
Well, that’s a pity for them as they won’t see any more money from me, if that turns out to be true.
They remove an essential functionality and then afterwards tell me “sorry, bad luck”? That for me is a trust problem which they are not going to recover from in my book.
My laptop is still working and I can use it with the hibernate workaround to circumvent the battery drain issues. But the fiddling required to just get the most fundamental thing to work on this hardware is a waste of lifetime for me at this point.
Maybe I am just going to print a case and use the board as a homelab server (that doesn’t require any sleep functionality) or hand this thing to my wife replacing her x230 as a “couch surfing” machine that’s connected to power most of the time anyway and go back to my MacBook.
As much as I love the framework idea, I hate being kicked in the nuts after 3 years being left with less than before. Good luck to everyone - updating BIOS can leave you with a handicapped laptop with no way back - good work framework… 
Which of the Windows states s0, (. . . . s3 s4 etc.) do you consider ‘deep’
I use hibernate but have just enabled s3 to see what happens.
It uses about 1w
@lvdd Did you try to use nvme.noacpi=1 as boot parameter? It drastically reduced the battery drain in s2idle for me to a similar level as deep (~ 0.5%/h).
Which of the Windows states s0, (. . . . s3 s4 etc.) do you consider ‘deep’
I don’t know the Windows states, but it should be s3.
0.5% of what per hour?
0.5% of battery charge level on a 55Wh battery would give 54/48 or just over 1.125 W/h
So my s3 is the same roughly at 1W/h
As for the Win states; Quoted July 2024
I’m stubborn and still trying to figure out why both of my 11th gen systems suspend and resume with no issues. Poor @lvdd ran through about eleventybillion different scenarios and didn’t get anywhere with determining why his system does not come back. It feels like a driver not reloading properly, but I am way out over my skis on this. I am just a dumb end user, tinkering and plinking away on the machine. I would love it if we could get an engineer (not Support - nothing against Support, but this is out of their realm as well) that would work with us to test working and non-working systems to try to narrow things down.
Damn, so much friction just to even ‘attempt’ to get it to sleep (and wake) right. (successfully or not).
@lvdd what distribution/kernel are you using? And anything else you can comment about your setup/software configuration/kernel parameters?
Hi Kieran,
thanks for chiming in.
There are a couple of links I posted with information about my system and the systemd journal during the death-loop. Also there is a description there how to reproduce it - at least on my laptop. The distribution doesn’t matter as I can easily reproduce it with Fedora 40, Debian stable and Archlinux.
The process for me is always the same:
Install a base system with Gnome (for convenience)
Add mem_sleep_default=deep via grub to the commandline
put the laptop to sleep via powerbutton
It goes to sleep but is not waking up anymore with the effect I have outlined here: Sleep issues after 3.19 BIOS update - 11th Gen Intel Framework 13 - #28 by lvdd
The point here is that this happened after I upgraded the BIOS from 3.17 to 3.20. All of the above was working fine before. Which lets me believe that this is not an issue with Linux Kernel changes but something being introduced with the BIOS.
If I could somehow go back to 3.17 I would do that and never touch BIOS again.
@lvdd for the bios update to 3.20 we updated the core bios kernel from Insyde, and CSME, and I suspect a change in there may have broken this somehow. Your support request got to me, and I was the one who gave the statement that S3 sleep is not supported.
This is actually the official stance of Intel supported sleep states from Intel. So even though it previously did work is was not an officially supported configuration.
I am not sure exactly where this got broken, but other vendors have documented similar issues with removed/missing S3 sleep, and the requirement for CSME to support S0ix sleep. Eg https://www.reddit.com/r/System76/comments/k7xrtz/comment/gev9mkp/?utm_source=reddit&utm_medium=web2x&context=3
We can do a quick pass to see if there was something that changed which caused this to be removed, however this may be something in CSME or other components which is not fixable by us. So no guarantees if this can be fixed, and no set timeline. (I would continue to operate under the assumption that this is not going to be fixed for now).
I would have to go back and measure again for this release, but we have generally found that Linux does a much better job with S0ix sleep power drain than windows. Eg on older intel platforms I think we were ball-parking around 300mW for Windows, and around 260mW for Linux.
Regarding that may be true the s3 doesn’t actually work as a quick wake up, it’s slower than hibernate and logs me out, so I’m back to s0, which works fine but does use watts of power.
Yes, I can attest to this.