S3 is unfortunately pretty much dead all-around, but S0ix continues to improve. We have firmware optimizations going in on both the system side and the Expansion Card side to help reduce bad retimer behavior around suspend though which improves power draw.

I think we talked about that before, but lmk if you need help with S0.

It takes some work, but it's possible to get to less than 0.5%/battery/hour, a fair benchmark as a battery reduced by >=5% after 10h of sleep is generally acceptable

holy crap, I'd love usage numbers that low. I'm running PopOS on a thinkpad, not on a framework, but S0 has been a long fight for me and I haven't felt like I've made much progress.

Do you have links to useful resources??

I have more experience with thinkpads, so yes ofc!

First please read my windows guide on csdvrx.github.io: it's very generic and should get you a baseline in Windows

Then get at least a week worth of measurements over long periods of sleep (like at night, reboot on windows) as a baseline

Then tell me how it compares to what you get on Linux after enabling everythig with powertop tunables and we'll work from there, as it'll be more precise than powertop.

Just send that by email at my outlook address. Includes the discharge line plot from Windows sleepstudy you find the more relevant, and a dmesg from linux

FYI, without wifi, on my thinkpad recently installed to Arch I measure 3.6W in powertop but I think it needs a recalibration as I measure 2W on a USBC power meter when idling with wifi enabled and a full wayland desktop (edge, etc)

> Has AMD introduced Apple-like sleep/wake/suspend behavior?

You have been able to configure that behavior (sleep-then-hibernate) on any Linux system that supports S3 for a long time (years and years).

At the same time, Apple now uses a kind of sleep similar to 'Windows Modern Standby'/Intel S0ix: https://macreports.com/why-does-my-mac-get-notifications-whi...