Nvidia dGPU blocks modern standby on laptop

@gorgor

Well, S0 sleep was going OK for a day or two with it consistently sleeping with <700mW discharge rate. However, it seemed to get worse the longer it had been since my last reboot.

I uploaded a PDF of my sleepstudy here: Dell XPS 9570 - Sleep Study after reinstall Win10 2004 in S01.pdf - Google Drive

It did seem that if I waited 30-45 seconds after unplugging my dock the computer would sleep better than if I didn't wait to put it to sleep after unplugging the dock.

In the end, I decided that S0 wasn't going to cut it for me. I hadn't even gotten to installing Hyper-V or Docker and I wasn't satisfied with the state of the sleeping. I researched the rEFInd w/ AcpiPatcher.efi solution that you referenced and came to the conclusion that both the people behind rEFInd and AcpiPatcher.efi are legitimate, so I decided to trust them. FYI, here's the discussion that lead to the AcpiPatcher.efi: Does not boot on Dell XPS 9570 . Issue #8 . datasone/grub-mod-setup_var . GitHub .

I switched to that solution a little bit ago and my computer seems to sleep fine now using S3 sleep. I'm hoping this works for a long time to come. I might see if I can keep Secure Boot enabled by adding signing keys to the Dell BIOS or implementing the shim-signed solution (of which I still am unclear of the process after some reading).

That guy in the post with datasone is me... and the Reddit post is also by me...

And yes, modern standby is a total disgrace. It wont ever work reliable. I cant speak for newer XPS laptops, maybe the 9570 has some fundamental issue which isnt fixable, or Dell being just incompetent as they usually are. But even if it "works", the drain is 3x larger compared to S3, and it just wakes up 2 seconds faster, which is just unbelievably STUPID. You treat 2 seconds faster wake up for 3 x worse drain and high chance to blow up, and melt your laptop. You can read my rage about it on the reddit post too in the comments.

The sad thing is, that there is some bug in the 9570 bios or EC, causing a bug when using S3 sleep, you can read about it here in my old post too:

https://www.reddit.com/r/Dell/commen...ep_and_modern/

That was always a reason for me, that S3 was not an option on the 9570. It is kinda ironic, that both arent working properly for the 9570 under Windows. Neither modern standby, or S3.

The 9570 was mostly Dells first "try" with modern standby, and they ruined something with it. I havent seen sleep problem reports so far with the new 9500 and 9700 line, lots of issues though with the 7590 of course too:

https://www.reddit.com/r/Dell/commen...hot_overnight/

I am 99% sure the 9500 and 9700 also have the issue, but people are not aware about it (yet). If I had access to a model, I would totally know in a week or so testing.

I made a ticket at Nvidia about this btw, they told me they send it to their developers and they have a look into it, I doubt they will find anything though or just tell me in the end, that they couldnt reproduce it.

It is just a pure joke, that MS doesnt offer a simple "sleep configure gui" in windows 10, where people can select either S3 or modern standby, and then also select something like "no wake timers" "no wake devices allowed", and then also have some "force all off" toggle, where no device and no software is ever allowed to wake the device with modern standby.

I mean how incompetent can you be? MS doesnt seem to know, that people use Windows 10 with laptops and tablets? They even make Surface devices which also have the same issues.

Modern standby is dead under Windows, and MS is too stubborn to acknowledge it.

@gorgor, Ha! Fancy that. :) Thanks for your help getting S3 back in Windows 10 2004. I can tell you've spent MANY hours trying to get to a resolution on this.

I've seen the C-states bug when coming out of S3, but that's manageable to me whereas huge battery drain while in S0 is not. Another thing that may need tested is if we do a clean install of Windows with the S3 from install time, will the C-states bug go away? Maybe that's a side-affect of switching between S0 and S3 sleep states without reinstalling Windows. Anyway, just a path to try that I'm not sure has been tested.

updated my post look above.

this c state bug I mean isnt a real c state bug, it is more a runaware issue on one core, after waking up from S3. My theory is, that this is somehow related to the EC and usb, because if you disconnect all USB, that also includes Bluetooth, the drain goes away after a few minutes. No it has nothing to do with that, I think, it must be some bug in the Dell EC/Bios. Dell or the company making the Dell bioses are incompetent. Read about the c state issue also here:

https://www.reddit.com/r/Dell/commen...on_the_xps_15/

This bug introduced with 1.14 was fixable I found out by pure luck, by loading bios factory defaults. A similar bug was also present before, where VT wasnt working anymore, and also fixable this way.

This is all connected in my theory, and the bios of most Dell laptops have this issue, the same is true for this toggle bug with undervolt, and also c state. I could explain this in depth more but it is a long story.

Might also be this is all connected with Intels MC updates for spectre and Meltdown, causing some incompatibility with Dell laptops bios.

If youre bored, you can test that theory, reinstall Windows 10 with refind patch present, so Windows installs in pure S3 mode, but I doubt that will fix the issue.

- - - Updated - - -

Also read this reddit topic, it is really interesting:

https://www.reddit.com/r/Dell/commen...cstates_after/

This seems to be some nasty bug, which is present in most Dell laptops bioses, mostly causing this, since over a year or so now and manifesting itself differently on different models.