Requests says nothing as well, sadly. Before it would sometimes say the UsoWorker service was waiting (I had registry disabled windows update so maybe that was just spinning forever and preventing sleep but I’ve long since re-enabled windows update to try and troubleshoot this)
LTT made a video on YouTube describing a similar behavior.
It seems like that the machine cannot detect power-state changed while it is asleep, thus if you plug your machine in it will maintain “network connected” state and screw everything up
Need to verify but seems promising.
1 Like
I just saw that video yesterday, It’s my exact issue!
Unfortunate that it’s not a framework specific thing. What’s worse is S3 doesn’t play nice with my eGPU, and because S0 and S3 are mutually exclusive based on a registry key and a restart, I can’t create a service to swap between the two on power status change. I’ll just have to live with it until either windows or Intel changes something I guess.
2 Likes
Looks like this video from LTT might be relevant to the Framework laptop, but not just the FW, but all Windows laptops, as Xavier_Jiang mentions here:
Hopefully this shows how this isn’t just a Framework thing.
The irony? This problem is why I ditched Windows 10 on my FW. I went to Linux because there this modern standby garbage doesn’t exist. And here is the thing, even if modern standby worked flawlessly, I have a problem with me telling my computer to sleep, and it doing what it wants to do instead.
I get that the average computer user is clueless, but this is a precedent that should not be embraced.
Anwyay, I’m really happy to hear that FW implemented S3 in their 12th gen bios!!!
6 Likes
Can confirm it’s available in 11th gen as well through the PlatformAoAcOverride=0 registry hack and a reboot.
2 Likes
For me, resume from S3 takes longer (12-13s) than from hibernate (10s), defeating the purpose.
As seen from early in this thread, other laptops drain a fraction of the power in s0ix vs. Framework. There is still a lot of room for FW-specific improvement.
Oh yes! Suspend to RAM ftw
Even on the Thinkpad T14 you would think that “Linux sleep” will give me support for S3
Not true. But, at least after I press the power button, the machine will go to sleep. Provided that I do not plug/unplug anything (including power)
Well, all the diaplay adapter cards contain active circuitry, so they will drain some power.
And then come the fact that we have slots for RAM, which is more inefficient than direct routing to soldered chips.
It also does not bypass the fact that S0IC/IX is absolute horrific.
It also depend on device manufacturers, and my Thinkpad seem to have a feature called “auto-hibernate” which will put the device into hybernation after sleep(ing) for a certain amount of time or after a (system) specified amount of power is drained. It’s called “dozing off” or something. I saw a log saying that and was super-confused.
You might have to connect/disconnect it before/after you sleep your computer.
Really? Although it does make some sense, since S3 turns off the CPU (mostly), which probably also includes the onboard Thunderbolt controllers.
That’s precisely what I want to avoid, haha. What happens if I try to enter S3 with it connected is that it disconnects and I have to unplug+replug to reconnect it.
I’d copy my custom service to handle brightness on/off AC and create one to swap the sleep type on AC attach (I only ever use my egpu when charging), but you need a restart after the registry hack to change from modern standby to S3.
1 Like
I’m a bit confused by the S3 sleep behavior, why does it take this long? On my X200 it wakes within 5s or so! Would it be possible for this to be patched through bios?
Although the people who use S3 is probably very limited considering it’s a mod on Windows (unless Linux uses it by default, haven’t checked).
S3 on WIndows under the 11th gen mainboard didn’t take long to resume. However there weren’t any real power savings from it that I noticed, so I switched back to S0 but with network disconnected.
2 Likes
S3 is “suspend to RAM” which basically turns off the CPU but keep the memory alive.
S0 Low Power keeps the CPU in a “low powered state”, which, even in S0 dont consume that many power (as I am typing my Ryzen 5 4650U reports 2.5-3W of power consumption) in the first place.
So depending on optimizations, chipset and OS routines it might take a while. Not to mention how Windows 11 is vastly different (from Windows 10).
Yeah I am aware of that, I was commenting on it not taking long to resume because others had questioned or mentioned that S3 took a long time to resume from. I think they might have been thinking of a bug under Linux that caused S3 to take a long time to resume from when running Linux and a specific kernel. (Long time here being a few seconds instead of instant.)
1 Like
S3 is also relegated to “linux/legacy sleep state”, which I guess is what that is causing the issue.
I think it’s down the not-so-great support state for S3 sleep, as covered in one of Linus’ more recent videos.
How long is that?
I use hibernate which takes 7 seconds to save to ‘NVMe’
On my dual boot which brings up GRUB from hibernate I have the 7 seconds to load from NVMe ’ select ‘Windows’ another 2 seconds i.e. 9 seconds before password request.
S0 was a mess and took at least as long as hibernate
Windows Suspend (that’s sleep ~ right) a different matter with it’s ‘instant’ showing without having to log in. Is there not meant to be a login when waking from sleep?
So I haven’t had the issue in either of my laptops since I have stopped unplugging the laptop when it’s sleeping. It’s kind of annoying because I have to wake up the laptop before unplugging it and then put it back to sleep.
1 Like
I’d like to see where my Framework 12th-gen stands on this issue - I’ve experienced some of the same problems as many other users.
However this problem’s reporting methodology among users is all over the map, understandably. Perhaps it would be good to make a suite of specific scenarios and battery run-down tools we can all try so we can get consistent information?
I’d be happy to spend some time going through specific steps, even if it takes a long time, but only if there’s a way to aggregate such data into something useful for troubleshooting. Currently poring through the comments is a frustrating experience, it’s hard to know what is relevant, and hard to compare results to other users.
I’m happy to help with this but I’d be of limited use, it’s out of my area of expertise.
[I’m not very active here in the forums, apologies if this isn’t the right place to put this comment]
Cross-posting this solution here because it has completely stopped random fan ramp-ups when in sleep from turning the inside of my laptop bag into a furnace. Laptop Fan Ramping Up in Sleep - #15 by Mr_Darcy
@BJ_Kramer Would you be willing to test this on your 12th gen? I can confirm this fixes the issue on 11th gen i5, Windows 10 Pro.
1 Like
While it’s still not perfect, I recently discovered that Windows Phone Link is responsible for a lot of CPU usage in sleep. With Phone Link running and the laptop on a charger, as soon as it goes to sleep it starts overheating with the fans blaring. With Phone Link turned off, same setup, there is no fan noise at all.
3 Likes
@Be_Far that seems to have worked for me to prevent it from overheating and fans turning on.
I’m still not getting proper sleep though, and the culprit is the audio controller. Did anyone find a proper solution to this?
1 Like
Update:
Not sure if any of these issues are framework specific, but documenting anyway for anyone in the future.
- Disabling Fx Sound removed the Intel SST problem (so far) and reduced power consumption to about a third
- There was a weird temporary issue with find my device, only happened once
- Quitting Phone link reduced power usage by ~20%, but it’s not something I’d want to do regularly.
- Main culprit now is the C10.Unkown, which I believe some people mentioned earlier
I’ve managed to get it down to about 2%/hour, down from 12%/hour, which is acceptable. If anyone has any advice about C10 let me know.
2 Likes