XPS 15 9570 Owners Thread

Short version, download this and follow the instructions

https://www.wagnardsoft.com/forums/viewtopic.php?f=5&t=1174

You might want to get rid of the Dell Intel GPU drivers as well while you are at if they are installed.

 

Thank you! So after uninstalling with DDU, should I install Geforce Experience from Nvidias homepage?

 

Yes, something like that, but the readme file is really detailed, just follow it and it should be fine.

 

Just so you know Geforce Experience doesn't contain the drivers.

You can get the Nvidia drivers from: https://www.geforce.com/drivers

This also gives you the option to install Geforce Experience while doing the driver installation. If no drivers are installed prior to trying to install Geforce Experience then it will not be able to find your video card.

 

Is anyone still having problems with the Killer wifi card? Haven't experienced much problems with mine, mostly it's slow to re-connect when I use VPN and pc goes into sleep/hibernate. But otherwise it always reconnects

 

Maybe I am remembering wrong here but I recall that Geforce Experience can download and install drivers as well?

 

The recommended procedure with DDU and reinstalling the Nvidia driver worked! When you install the driver you also get a question about installing Geforce Experience so everything seems to be solved. Thank you!

 

@Wildfire12 Glad to hear that you got it sorted

 

It will download and install the latest drivers but only if you have the hardware already identified on the system (drivers installed). I don't believe you can install Geforce Experience with the devices flagged in device manager.

 

You could be right here, actually don't remember how I did it last time, maybe I used the drivers GE had already downloaded after running DDU.

Sent from my SM-N950F using Tapatalk

 

I reached out to dell and they effectively told me that the sleep issues are Window's problem (which is probably true). Doing some google searches, it does seem even Surface's are not a stranger to this issue.

Also, since the new bios, i've hard a weird issue where when I start a game up, like WoW, it takes about a minute for the system to kick into high gear and use faster clocks. Anyone have this issue or have suggestions?

 

The problem IS MODERN STANDBY...

^

I have an interesting information to share though on this. I bought a Samsung 970 EVO to replace my SKH. So I also did a clean Windows install. I wanted to try to debug step by step, if maybe some driver causes the c state bug with S3. So I installed Windows, and installed no drivers but chipset and GPU and then deactivated modern standby via registry and monitored c states with TS... and the resulting is the following:

After a fresh Windows 10 install with just chipset drivers, thermal drivers and GPU... S3 works normally, no c state bug. S3 works beautiful, c states work normally after wake up. I was totally happy and began to reinstall all other drivers, always monitored c states again after each reboot and did a S3 sleep... all working, no c2 stuck bug. I began to think this may be resolved. Then 2 hours later... I checked again, and now it is broken again and won't work anymore, whatever I do. C would be stuck in C2 after S3.

I have no idea why. Something changed in that 3 hours and I have no idea what and why it could cause this. I just changed from local account to MS account. I have some suspicion that Windows 10 installed a bunch of drivers over Windows update in that 3 hours of idle time because I see that in the Windows logs, even I deactivated automatic driver install via OOSU as one of the first things before Wifi driver install... which seems to be ignored, which is just COOOL. Totally COOL. I googled about it and even if you deactivate auto driver update, Windows 10 still will do it for most drivers (except GPU, network, and some more, but "important" drivers it will just update regardless of deactivating driver update), so you cant prevent that. And it seems something it installed in the background is causing the c state bug.

If anyone want to help to figure out, what the culprit is of the bug, I would be glad about it. Maybe someone here wanted to do a fresh Windows 10 install anyway and can help to debug step by step which could lead to the broken state.

Because... S3 worked perfectly after deactivating modern standby via registry in the first 2 hours of fresh Windows install. THIS is why I hate Windows 10 so much. Because of COUNTLESS of tiny little bugs like this, the normal user has no idea about, which just lurk in the system.

I am sure the issue was not triggered by a normal Windows 10 update because I did that, and checked after reboot, c states worked still. It must be something else Windows 10 did afterward.

The Intel Audio Service wasnt also crashing in the beginning, Windows 10 installed a driver through store which then caused it.

The stuck at Dell logo also randomly happens now again

I will try another Windows 10 install and wont install Wifi drivers, so it cant do Windows Upates, install all Dell drivers and then look if c state works or not, so it must be some Windows update / driver through Windows update causing it.

 

Maybe Windows needs few hours to settle in and enable some hidden internal setting that messes up the S3?
Could you try to make a longer pauses between driver updates, like a day after each driver update, and see if the sleep still works? That should give the sufficient time to know if any of those driver updates messes up the sleep...
Or even not install ANY driver (but give it wifi) and see if the sleep will get messed up within a day...

Is it possible to enforce the S3 sleep via registry only with the old BIOS 1.2.2 or also with the new BIOS 1.3.0 that disabled the S3?

 

Ok I did another Windows install... and I was able to identify the cause of the c state bug. Well, kinda. In other words, I found a workaround how to prevent it and indirectly the cause of it. It seems to be a Windows bug, but mostly also in combination with Dell bios bug, because, I dont see any reports of this on the web with other laptops. I dont have another laptop to test this though, which uses modern standby. Maybe someone with a 2018 model can test this, if it is a general Windows 10 bug, or a Dell XPS 15 bios bug.

Anyway... the cause of the c state bug is: Pure S3 (suspend to RAM). I found by luck, that if you disable (or let Windows disable it) hybrid sleep, then the c state bug will happen. If you keep hybrid sleep enabled, it wont happen and WILL work normally. What is hybrid sleep? It is basically a mix of S3 and hibernate. So you willl basically have S3 (fast wakeup), but Windows will also write "some" data (not clear out of MS docs what data, it doesnt seem to be the whole RAM) to disk as a "backup" if S3 fails (battery empty or power cut).

This is totally weird though, because, S3 is still functioning normally. I have no idea why a "pure S3" triggers the c state bug and the hybrid sleep not. Maybe in hybrid sleep, Windows writes the kernel memory to disk and loads it then every time when waking up meaning it "restarts" the CPU maybe, and this prevents this c state bug.

So... if you let hybrid sleep enabled... you will have WORKING s3, no c state bug, fast wakeups, safe sleep, little drain, laptop totally off during sleep. The only "disadvantage" now is, that you will have a bit of write to your SSD everytime you put it to sleep. If a normal SSD has a lifetime of maybe around 500TB write, thats mostly negligible.

So how to get this?

1. Start the registry editor (regedit.exe).
2. Move to HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Power.
3. Double-click CsEnabled and set to 0, then click OK.

Reboot and done. Modern standby is now deactivated, Windows 10 will use S3 if you put it to sleep. You have to check if hybrid sleep is enabled for battery and AC under energy settings, otherwise, it will trigger the c state bug after waking up. This works regardless of the bios update to 1.3.x. You don't need to reinstall Windows. You should select the Windows balanced energy profile though, not the Dell one if you have one.

You will need to change CsEnabled to 0 after every major Windows update, that are the creators/fall/spring updates happening 1-2 times per year, because the registry is set to default by that.

 

I couldn't see what size 970 Evo you installed from your posts, but here's the spec from Samsung that shows the TBW per drive.

How much is written each time you put the machine to sleep, roughly?

* TBW: Terabytes Written
* Warrantied TBW for 970 EVO: 150 TBW for 250GB model, 300 TBW for 500GB model, 600 TBW for 1TB model, 1,200 TBW for 2TB model.

 

@Woodking I said already 500TBW as a middle. Even if you assume a worst case of 32GBW per day, that would be 12TBW per year, meaning the drive may fail after 50 years...

 

@maffle Is the bug you're referring to where the CPU stays at like 2-3W on idle instead of 0.7-1.2W?

 

Yes.

 

This sounds great if it works! I'll try it.
But I don't have any setting for hybrid sleep in my Balanced (or Dell) advanced power settings, only the basic Sleep after and Hibernate after settings. How to enable Hybrid sleep if it's not displayed in the settings?

 

What is the verdict on the 1080p vs 4K screen on this? Noticed in UK they are now providing 97wh 1080p versions, which is tempting.

 

... you obviously didnt deactivate modern standby (correctly) first in registry and rebooted.

 

Ok. Yes, this is infuriating. I have seen it on other laptops as well (x1 carbon)

 

Where do you get the figure of 32GBW per day, or do you surely mean 32GBW per time you use sleep? Starts adding up if you constantly use sleep.

 

Nice findings!
I am sorry to curb your enthusiasm but...
My understanding of hybrid sleep is that it keeps the ram powered (like S3) AND it saves the whole RAM on disk (just like hibernate), not a tiny fraction.
If you have a power failure while you are in hybrid sleep, and hence you lose the RAM content, the system is completely resumed from the disk's dump of the RAM (this has happened to me on my desktop).
Hybrid sleep is default ON on desktops and default OFF on laptops.

https://www.maketecheasier.com/hybrid-sleep-on-windows/
https://www.howtogeek.com/102897/whats-the-difference-between-sleep-and-hibernate-in-windows/
https://blogs.msdn.microsoft.com/oldnewthing/20110510-00/?p=10703

 

In my case roughly 2GB each time. I have 20GB installed.
I use hwinfo to check disk before and after hybrid sleep

 

Actual disk writing is less than the memory size. They compress the the memory and do some other smart stuff to limit the file size

 

Well of course when I say the whole RAM, I mean the amount of memory currently in use. No point in saving zeros. And is probably compressed as well.
However, these things are also true for the regular hibernate.

what I am trying to say is that the amount of data written on disk is the same for both hybrid sleep and hibernate.

 

That was a worst-case scenario like I said. Just a worst-case calculation to show, EVEN IF, you/it writes 32GB (the whole RAM if you have 32GB installed), what it doesn't or in general mostly never does, because you mostly never have the 32GB in use and 32GB uncompressible data in RAM, the life time of a normal 500GB SSD is then about 50 years for a 32GBW per day. So in general, if it maybe writes about 2-4GB per sleep activation, and you put it to sleep a few times per day to sleep, depends on your use case, then even it may be more than 50 years.

I knew that, and it was already intentionally thought of in what I said. So even if hybrid is 1:1 the same amount as true hibernate of written data, it is the only working alternative now as I see it. And if you think about it, with the numbers I said above, does it matter? Even if you are "generous" and say Windows writes 64GB of data to SSD per day, the SSD may fail after 25 years.

That is what I expected too.

So... as I see it, we have the following options now with the XPS:

Modern Standby
- fastest wake ups (0.5-1second)
- terrible drain (1-3% per hour), high risk of a catastrophic drain (3-5% per hour)
- not reliable
- dangerous (laptop can burn in bag)

S3
- fast wake ups (1-2 seconds)
- reliable
- smallest possible drain (0.1-0.3% per hour)
- unfortunate right now triggers C state bug (c2 lowest c state, idle package power of 1-2W instead of 0.5W)

Hybrid Sleep
- same positives as S3
- used RAM is written compressed to disk
- no c state bug

Hibernate
- slow wake ups (5-10 seconds)
- same negatives as hybrid sleep (disk write)
- reliable

So the only option as I see it, is hybrid sleep, and I will use it from now on. I will somehow look into trying to contact Microsoft though and report the c state bug, maybe someone here has a suggestion how to do it. Because... as I see it, there is no good way to contact MS with a good chance they care and the report gets an "open ear". They mostly wont listen to it and care anyway, because they have their holy grail modern standby, which is unfortunate, off the track, and wont ever work in Windows in my opinion.
Why does it work in iOS? Well, because iOS doesnt allow apps and processes to run in the background that openly like Windows does, and iOS has strict drivers all controlled by Apple.

You can do the calculation for your SSD yourself, just look up the TBW. A normal 500GB SSD is roughly about 500TBW after it may fail. Just do the math how often YOU put your laptop to sleep per day, how much USED RAM you have open in general, ect. And you mostly will see, that even if you have some worst cast scenario sometimes or often, the SSD may fail after 25-50 years.

As I see it, hybrid sleep is the only option right now. Or you shutdown your laptop everytime you dont use it...

 

OK, so I set the CsEnabled registry key to 0, rebooted and got all the additional power settings. The hybrid sleep was already enabled.
I worked for some time on the laptop, and then closed it, with 5-6 apps open, wifi and BT on.
EDIT:
And in 3:20 hours the laptop lost around 1.1% of battery. It took 1-2 sec to wake up, a bit longer than the modern standby.
The powercfg /sleepstudy shows the laptop was in the "standby" (as opposed to "modern standby") the whole time,
So that's good - less battery drain and still a quick resume from sleep.

HOWEVER, now I'm getting a whole screen glitches literally once every minute, for just a split of a second, but over the whole screen. What's that about??
They look like the screen glitches I was getting on my old Yoga 2 Pro when the Panel self refresh option was enabled in Intel graphics settings. But the panel self refresh is disabled (it's disabled because I have the keyboard skipping bug, and that's an "official" workaround until the BIOS fix is released).

 

I probably will continue use S3 (BIOS 1.2). C state bug is annoying but is not so critical to me, since if the laptop is on and on battery I will be certainly doing something.

I am not too concern on hard drive write. I have a 1TB drive with endurance of 300 TBW. and i will be very happy if that one last for 5 years and by that time 1TB ssd should be dirt cheap.

 

No offense but it wasn't intentionally written in what you said before.

You were trying to minimize its impact. Because you are self-convincing yourself that it is OK to dump your RAM in your in your SSD every time you put the laptop to sleep while you were ranting about it several pages ago.

Anyway, I do hope your hatred for this bug will push you to annoy Microsoft enough so they can fix it.

And BTW, while S3 sleep consumes less power, but it's not 10x better. From my experiments, it's probably close to 2x.

 

I have been having this screen glitches for some time. Generally if you keep on doing standby/wakeup eventually there is a chance that this screen glitch starts showing up. And once it is up the only way you could get rid of that is to do a full restart of the system.
More recently I removed intel graphics driver downloaded from Dell and install the latest driver from Intel. So far I have not seen any but it is a probability thing so time will tell.

 

Just a friendly reminder to keep things civil. It's great to discuss the topic but I'll be keeping an eye on this thread...

http://forum.notebookreview.com/threads/forum-rules.109941/
While debating and discussion is fine, we will not tolerate rudeness, insulting posts, personal attacks or purposeless inflammatory posts. Our decision is final in these matters.

 

on my system I am seeing a 4x difference. Modern standby 2% per hour, S3 is about 1% per 2 hours.

 

Well, that''s not good. I never had it until I let the laptop sleep with the S3 sleep for the first time. Looks like it's related...
I enabled the panel self refresh, glitches didn't stop. And then I disabled it again, and they stopped, or there were just couple of them and they stopped. Let's see how it goes...

EDIT (I made a mistake):
I just had a 1.1% battery drain for 3:20 hours, so that's around 3 times less than the usual modern standby power drain for the same time.

However, I'm seeing a higher idle power drain now, 2-3W more. And the TS disappeared from the system tray, but it's still in the memory. But I can't open it from the tray any more...
How to check for the C2 state bug?

 

2-3W idle power drain is the C2 state bug.

 

I noticed that too, but I had similar glitches with modern standby a few times, even had black screens after opening the lid. I thought the laptop crashed but it didnt. After pressing power key for 5 secs, Windows actually "scrolled up" with a are you sure you want to power off with unsaved documents open? And the picture got black... I had similar glitches before on my old Broadwell y tablet (Dell Venue 7140). This is why Windows 10 is just not usable. These bugs exist since years and didnt get fixed so far. I will monitor for glitches. I could actually see now may be a bug with the Intel drivers, after waking up from S3, that would be just fantastic.

You can update the bios to 1.3.1, and then just deactivate modern standby via registry, thats basically the same as via bios, so it wont matter Dell took that option away from bios. The one option is s3 force, the other "OS", so Windows decides, and then you can disable it via registry.

A 3-4x worse drain seems typical, FOR """WORKING""" modern standby. But if HW deep sleep drops by some glitches which just randomly happen all the time, it gets worse, way worse up to 10 times worse, so that is totally correct what I said before. Even if it """just""" 3x... it is a difference if it drained 10% or 30% ... right??? For doing NOTHING.... NOTHING.

Do you have put TS into your task scheduler? If so, there is a stupid bug, you have to deactivate all the grayed out checkboxes in the options, otherwise, Windows will close TS if you switch to standby.




There is also a bug with Windows 10's systray (how can it be different...). You have to enable "show all systray" icons to prevent half of them "vanish" randomly. This bug is present since over 1 year now and MS doesnt care. It is discussed here: https://answers.microsoft.com/en-us...essageId=38b484eb-93e8-4056-951f-61dcd1482396

And again, these things make Windows 10 just not usable.

You can check c states with TS. If you have 2-3W idle package power, Im sure there is something else in C0 causing it, look in your task manager. Also check if your GPU1 is off, maybe the glitches are related to it and it turns on the Nvidia GPU randomly.

This is how it should be in idle, and how it is for me after hybrid sleep and modern standby without the c state bug.

press c10 button and monitor c states, this is how it should be:

if its bugged, it is 100% in c2, instead of c8 or c10.

 

Anyone got a picture of the VRMs on the 9570 by chance?

 

Any more from you and you'll be on the cricket list!

 

At least in previos XPS15 version, the FHD screen has been among the best in the business and this year they are supposed to have increased the brightness as well. If that option (big battery + FHD) had been available when I ordered, I might have prefered that one to the 4K version. That said, the 4K panel itself is great, just that 4K on 15"...

 

You would think with the 30 pictures I took that one or two would have clear images of the VRM but sadly that's not the case.

This video shows some of the VRM around 4:50 and probably other spots too:

 

OK, thanks a lot for all the info.
With the S3 sleep and the hybrid sleep option, my package power in TS now varies between 2.4 and 4.5W. It was the same as yours, 0.4W on the modern standby.
And in C states in package C states it shows C2 takes 90-95% of time, every other C state except the Total is at 0%.
So I suppose this a a clear C2 bug. :-(

I guess I'll have to go back to the modern standby...
Pity - my laptop just had 11 hours of sleep and 3.5% of battery drain with the old S3 sleep.

But the idle power drain is much more important for me and I anyway use the sleep just for 2-4 hours max, so that's 5% in the worst case, or just 2% in the best case, which is not a problem. And I haven't had any situations where the laptop would wake up from the modern standby. In fact, I could see in the event viewer that it always shows something like "screen suppression detected" (or similar), just before the laptop goes back to sleep after being woken up for 10-20 sec. So I think the modern standby detects if the lid is closed, and will put the laptop back to sleep if it wakes up with the lid closed.

@maffle if your laptop doesn't have the C2 state bug with the S3 sleep then maybe this whole S3 sleep problem has to do with some specific driver, rather than the hybrid sleep option?

 

Window registry hack will bring back hybrid sleep. For pure S3 you need further disable hibernation.
I prefer S3 sleep only, and hibernation when battery is low.
I feel Microsoft is more and more like Apple. MS used to give user choices, now they try to tell user what is the right (and only) way to do. Now the problem is that MS never has a product which is as polished as Apple

 

After thinking long and hard between Aero 15x, Razer Blade, XPS 9570 and MacBook, I think I am going to go for the XPS 9570 finally (for various reasons).

The model I will be getting is the i7 FHD 16GB RAM 1050 TI 512GB SSD model.

Now it seems these are steps I will likely need to take - please correct/advise if you can:

i. Do a full clean install of Windows
ii. Repaste CPU/GPU (any recommendations on video's and thermal pastes?)
iii. Undervolt CPU

Additionally, these are two things I am looking at doing

i. Replace Killer WiFi card with an Intel one (any video's/purchase sites?)
ii. Upgrading to 32 GB RAM (does 16GB model come with 1x 16GB or 2x 8GB?).

Also, I've been reading about some sleep/hibernation issues, is there any true fixes for it and how much of a problem is it?

 

No, youre confusing. You can just disable hybrid sleep, that was the whole point of where the c stuck came from... Just update bios and set CsEnabled to 0 and then deactivate hybrid sleep in energy settings and you have normal S3 and hibernate when you want. The option for hybrid sleep will pop up after reboot when you set CsEnabled to 0, together with lots of other energy profile settings, which are hidden by MS if modern standby is enabled.

I am sure you just did something wrong. Make sure you deactivate fast boot, reboot Windows. Make sure you have really hybrid sleep enabled for both ac and battery. Make sure you really use the correct profile where you set it. Make sure hibernate is enabled. If you triggered c state bug before you have to properly reboot the laptop. Make sure what powercfg /a says it will or can use (it has to say that it can use hybrid sleep). I am totally sure you used normal S3 and not hybrid sleep. You didnt understand the whole thing. Of course, mine has it with S3 but just with pure S3... not with hybrid which is S3 too. I tried to explain it before in that post. My theory is that Windows maybe loads the kernel from disk with hybrid (or something else) and reloads the CPU, where in pure S3 it is somehow stuck.

 

Has anyone ever gotten a Samsung SSD on their 9570?

Sent from my SM-N950F using Tapatalk

 

If you order a 250 or 500GB variant, you have no chance getting one, just if you order a model with 1TB+. But then the chances are really slim. I have seen some reports of Samsung SSD with 1TB, I cant confirm it of course though.

 

I have the Samsung 970 EVO 2 TB up and running on my 9570.

 

Well, I don't know what went wrong, but all the pints you mentioned were set up properly.
The hybrid sleep was on for both AC and battery, I used the Windows Balanced power plan, hibernation enabled, fast boot deactivated, and I both rebooted and powered off the laptop few times.
So I don't know what went wrong, but the CPU got stuck in C2 state.
The power usage when sleeping was good though - around 0.3 watts per hour, or less than 1 percent of battery for 3 hours.

Does a 2 TB or even a 1 TB SSD use more power on idle than let's say 500 GB SSD?

 

@splus whats the output of "powercfg /a" after you disabled modern standby and rebooted? also, monitor disk write with a SMART tool. does it increase after you put it to sleep by a decent value? if not, windows doesnt use hybrid sleep but just S3 alone.