XPS 15 9570 Owners Thread

@darkassain can't see the pictures, but I tend to agree with your hypothesis.

 

@abujafar just fixed it
forgot to mention both times i had a HDD connected to the USB3 port and have my MX Master with the dongle connected as well.

 

I am telling you now for the third time. You are mixing two separate issues here. You also dont seem to have re-read my post on this of the other day.

1. having something connected to TB => package c state will stay in c2/c3, that is normal as I see it, and not a bug. Mostly because of the PCH doesnt support energy saving for TB or needs to stay in c2/c3 for compatibility reasons. Could be wrong on this, but my guess is, the CPU has to stay in c2/c3 if you have anything connected to TB, and that is by design. A discharge of 4000 is not normal of course. So that is another bug, that it wont deactivate TB during sleep.

2. the stuck in c2 after S3 sleep has nothing to do with this issue, but you can get it unstuck with plugging in something in TB and back out. could be, that sleep doesnt properly reinitialize the TB controller though and this is related. though deactivating TB in bios doesnt help.

3. BUT, that wont fix the drain. because afterwards, the package power is still at 1.5W and there is some loop bug on core0 of about 12% C0%.

Read this:

http://forum.notebookreview.com/threads/xps-15-9570-owners-thread.817008/page-138#post-10799137 (the last part)
https://communities.intel.com/message/556639 (last 3 posts)

Other laptops actually have the ability in bios, to turn off things like USB and TB in sleep. The Dell bios lacks this option. There is no reason to have anything on or on power during sleep, especially not TB or USB devices. Leading back to, how modern standby is a failure. Windows 10 also doesnt allow you to deactivate USB HID devices (BT for example) during modern standby. And it is the same with TB it seems.

If you read through this as an example of Bluetooth sleep in modern standby

https://github.com/MicrosoftDocs/co...er-management-for-modern-standby-platforms.md

Windows 10 has the abilities to handle this, but... it is locked (for whatever reason) and you cant change wake up events, or toggle power off manually for modern standby, where you were able before, for S3. This is the whole thing with "connected standby" and MS doesnt want you to turn things off, during sleep.

For Bluetooth and the issue with the XPS, that BT mice and keyboards wake the laptop up from modern standby. As I see it and understand the documentary by MS, the issue is, that the BT card (Killer or Intel doesnt matter), is connected to the board via normal USB. The Intel 9260 card has a USB interface for its BT part as you can read on Intels site. So, the whole issue here again is, that the XPS doesnt disable power for USB devices during sleep, BT then automatically included.

I am not sure if this can be changed by Dell with a bios update, and allow power off for TB and USB during sleep or if it's a limitation of the mainboard design.

 

@maffle listen, I read your posts.
At the moment, nobody knows for sure what is going on. We detect some symptoms but we can't be certain if there is one or multiple underlying problems or if they are connected.
You have to understand that your diagnosis of the problems are just YOUR hypothesis and YOUR opinions.
Your posts are always imperative. There is never an "I think", "I believe", "it could be". No, you KNOW everything and we are a bunch of idiots.
Thank God we have you here to enlight us.

 

@abujafar listen, If you had read them, you wouldnt have said this:

"Can somebody who has throttlestop and a usb-c device, check if they get the "stuck in C2" when they connected something to the usb-c port?"

Because I was talking about THAT and way more days before, but you know that, right? You read the post.

 

sorry, for not adding "someone else".

 

Did a little more testing, and actually found out something interesting.

Scenario: S3 wake up (c2 bug) + USB3 usb dongle via USB-c adapter at TB port



The whole thing is a bit complicated to explain, if I could make a video, it would be easier to understand. But I will try to explain. If you go into device manager, and look at the second USB 3.1 host controller, which is the USB host controller of the TB port, and you go into energy saving tab, and deactivate "this computer can turn it off to save energy". The moment you do that, with a connected USB3 to USB-c adapter, the C8 state will go away, and it will go into c3. If you activate it again and wait about 20 seconds, it will go back into C8 state.

I also noticed, that if you open for example Explorer and go onto the USB3 dongle (via TB port) and browse through its content, the C8 state will also go back to C3 (close Explorer, wait 20 seconds, it goes back to C8). So this shows, that anything connected to the TB port and is in "access/use", will prevent a C8 state. That would be a TB dock mostly too, if it has some devices like HDMI output chip, network, any USB3 ports or cardreader and they are in use. This would explain the C2/C3 state if you have a TB dock connected.

I was also able, to actually get a C8 AND 0.5W package power back, totally by accident, by toggling back and forth energy saving in device manager of the host controller, and also sticking out the dongle a few times back and forth and deactivating/reactivating also the devices in device manager... suddenly, I had 0.5W package power, C8 state... no C0% usage anymore AFTER I had triggered a S3 sleep before AND with connected dongle. I plugged the dongle back out, in and out... and it was broken again.

I tried to reproduce this, but couldn't so far, and I have no idea why. Maybe someone being bored, could test and try to reproduce this and get an unbroken 0.5W state again after S3 and the above scenario of toggling back and forth energy saving/deactivating device and so on with a connected USB3 dongle with USB-c adapter on the TB port.

The whole thing seems to be some nasty driver bug with Windows 10.

 

What you are describing actually reinforces the idea that all these things (thunderbolt, sleep problems, c2 states) might be connected.
I think the core problem lies in the logic that manages the package c states.
I am not sure where it is implemented, though.
Is it something windows does? Is it something partially hardware handled by dell's firmware?

 

The c2 bug seems to happen so far just on Dell laptops, I didnt saw a report with other brands. I spoke to some people over on Lenovo forum, and asked them politely to test this on their X1 carbon laptop with TS, and they have no C2 stuck bug after S3 sleep. So, it seems to be a bug with Dell's bios or EC. Could also be a combination of driver and Dell bios.

I also had a stuck at Dell logo again, after I activated modern standby and rebooted. I wonder if this c2 stuck bug (c0% on core0, in the picture above there are actually now two cores stuck after I toggled the stick a lot in and out) could actually result in the yellow yellow white blink pattern and no boot.

 

I don't know about you guys, but I've had an extra battery drain (I think this is C2) after every reboot, which goes on until I put the laptop to sleep, even for 5 sec, and after coming out of sleep the drain is gone.
And interesting thing is that I've had this issue for several years already on my old Lenovo Yoga Haswell laptop with S3 sleep as well (with the same fix - put the laptop to sleep for 5 sec), so I think there's some Windows bug going on here, AS WELL (apart from the USB, thunderbolt and other issues).

I've noticed few times already that in the Event viewer I get a warning every 10 min DURING the sleep that an USB device is causing a drain and couldn't be unplugged because of recent I/O. Upon checking powercfg /sleepstudy, I could see the drain was coming from both Atheros bluetooth and Intel's smart sound driver.
I suppose this could just be drivers not doing its job properly (or just not talking to each other properly), and might be worth checking different drivers...

 

Dell just released BIOS 1.5.0 with few notable fixes. From the release notes:

Fixes:
- Updated Intel ME Firmware to address security advisories INTEL-SA-00125 (CVE-2018-3655), Intel-SA-00141 (CVE-2018-3657, CVE-2018-3658 and CVE-2018-3616) and Intel-SA-00131 (CVE-2018-3643 and CVE-2018-3644).
(Meltdown/Spectre/other security fixes? Will this slow down the CPU???)
- Improved sleep mode power management in Ubuntu. (could this be the return of S3??)
- Solved the intermittent issue where the system powers off from hibernation after adapter is plugged or unplugged.
- Fixed the issue with touch screen where it malfunctions after closing and opening the lid.
- Fixed the issue with missing characters while typing when system is in idle mode.

Enhancements:
- Supports wake from dock LAN feature.


Intel also today released a new display driver that works OK with the new Windows fall update, but among other additions and fixes, here are few:
- HDR and EDR enhancements for external and laptop displays
- Support for Wide Color Gamut Displays (relevant for XPS 4K display??)
- Improves hot plug audio synchronization between the Graphics and Audio drivers (could this fix the modern sleep power drain??)
- Improved color quality during video playback and battery lifetime when display is in standby
- Fixed: Panel Self Refresh remains disabled after full-screen video playback, which may impact battery life


Fingers crossed these 2 updates fix more than they break!!!

 

The CVE fixes are all Intel ME related, not spectre @splus

I just flashed 1.5.0 taking the risk, and I can confirm, C2 stuck after S3 sleep is semi fixed with 1.5.0. The CPU wont be stuck anymore in C2, and C8 works after S3. Yet, the abnormal 8-12% C0% on core0 is still present after S3 wakeup, resulting in a 1.5W drain instead of 0.5W for idle. " Improved sleep mode power management in Ubuntu" was the same phrase Lenovo used for giving back S3, because they dont want to use the word correctly with Windows which is "modern standy on paper".

Edit: I just found out, that removing the Logitech Unyfing dongle causes the C0%
to vanish after about 5-10 seconds, and you have 0.5W package power again... Doing now again S3 with no Logitech dongle, wake up, there is still a 12% C0% on core0 for about 20 seconds but then it vanishes from alone and you get 0.5W package power.

So if you dont have anything connected to the XPS, C2 bug seems to be fixed now with S3 sleep. I will test more on this.

 

I also updated and can confirm the 74C GPU throttling is still there. I can also confirm the downgrade flag is still set and you CANNOT go back to an older bios even with recovery mode.

I think I still have a bios dump from an earlier bios, perhaps I can compare the two and see if there are any easy to change GPU/S3 flags. I won't have time to look until tomorrow, but maybe we can fix what Dell broke.

 

Have you tried the latest BIOS (1.5.0)? Seems like they looked into it:

https://www.dell.com/support/home/u...W&osCode=WT64A&productCode=xps-15-9570-laptop

 

Lol, literally two posts above you. @maffle is looking into the S3/C2 stuff.

 

Any change in DPC latency? (specifically for ACPI.sys)

Oops, I just scrolled down on page 150 or so, and thought I was on the last page and didn't see the post. My bad.

 

Sadly in regards to DPC, I never ran it before. I can try tomorrow if no one else does, but I'm not holing my breath. Even my 12 core desktop seems to have issues with DPC latency.

 

Thanks, custom, that would be great. Have a look in the 'drivers' tab in LatencyMon and see what the highest execution time for ACPI.sys is after 30 seconds (should be enough).

 

No, not fixed.

I have a weird new issue now... I wanted to reconnect my MX2S via Bluetooth, because of the C0% issue after S3 with the Unifying dongle.... aaaand... I cant find any Bluetooth devices now anymore with the XPS. It totally worked some hours ago still, before I updated to 1.5.0. I have no idea how this can be related to the bios update. I tried to reinstall the Intel 9260 BT drivers, no luck. My Bluetooth keyboard still works connecting to it, but the "searching for new devices" show nothing anymore, cant find any devices anymore. And I also cant find it anymore with my cell phone for example, while the Windows 10 BT window is open.

Edit: Ok BT working again... I noticed the Bluetooth support service was set to deactivated under services, WTF @ Windows 10. Totally no idea, why this was the case, I totally didnt do that. And it worked before I went to 1.5.0 some hours ago, because I was always toggling back and forth my MX2S between unyfing and BT for testing scenarios...

 

Woke up this morning with exciting news, will try the new BIOS asap. It would be great If they reintroduced and fixed S3 sleep!

@custom90gt, I noticed in the BIOS download page that there is a message:
Important Information
Since 1.4.1 has Boot Guard updated, once updating to 1.4.1 or later BIOS, system can't roll back to the BIOS earlier than 1.4.1.

SO maybe this downgrade lock is not due to a flag in the BIOS.
I pretty sure that somewhere in the BIOS there are hidden variables/settings for S3 and maybe even the eGPU Temp Limit.

 

... doing some tests, now with BT, and I still had a 8-12% c0% on core0 (resulting in 1.5W package power, though c8 works now) after s3 wake up, which didnt go away. The first one was resolved the moment I removed the Logitech dongle. Now I had none connected. I then just deactivated BT and in that moment the c0% went away and I had 0.5W package power again. So there still seems to be some bug with USB after S3 wakeup, because BT is connected via USB on the XPS. The c2 stuck issue after S3 sleep is resolved though for me. Sigh... this will be fun, trying to explain to Dell CS. Anyone see the same?

 

Isn't BT on the same card as the WI-FI adapter, connected to the PCI-E bus?

 

No. Only the Wifi part of the card is connected via PCIe, the BT part of the card is connected via normal USB.

https://ark.intel.com/products/99445/Intel-Wireless-AC-9260



You can see that via a USB-Tree tool too ( https://www.uwe-sieber.de/usbtreeview_e.html ). Intel 9260 is the only new line Intel card, which still uses BT via USB, mostly because of backwards compatibility reasons for older laptops. The other Intel cards, like 9560 for example, use a CNVio Interface, which isnt compatible with the XPS. Thats the reason why only the 9260 is compatible to the XPS, as a replacement for the Killer card, which obviously also uses USB for BT.

" Intel Offers the CNVio Interface as a Substitute for PCIe and USB in Wireless Modules

SoC Intel Gemini Lake products, Cannon Lake processors, as well as Intel 300-series chipsets for desktop Coffee Lake CPUs will get builtin CNVi (Connectivity Integration) modules. It’s already clear, that these modules will contribute to Intel platforms wireless connectivity in Wi-Fi and Bluetooth networks."

 

I'm hoping they're catching up to issues and will address the GPU ceiling (assuming it's even a bug in their eyes =_=)

 

Just me having problems with Windows 10 constantly downgrading the Inte GPU drivers to an earlier version? And then Inte driver updater wants to update it...does...and then windows want to downgrade it...and so it continue.

Edit:
Solution seems to be here. But do it just after Intel GPU driver installation wants to restart or it will probably be to late.

https://support.microsoft.com/en-us/help/4026726/windows-hide-windows-updates-or-driver-updates

 

You can use OOSU for that. Everyone using Win10 should do that anyway for spying reasons of MS on you:

https://www.oo-software.com/en/shutup10

Then use "select advised options" in top menu, and then also scroll down to updates and select, "dont let windows update drivers" and "dont let windows download specific software for my device", close and reboot. It also deactivates all the hidden background telemetry services, which constantly spy on you and take away system resources, which is of course not wanted on a mobile device. Cortana, telemetry, feedback, tips, privacy options, ect.

If you have Nvidia GFE installed, also make sure to deactivate the inbuild overlay, because it constantly uses 1-2% cpu load if you dont. If you dont need it, I would remove it too because it also contains telemetry services spying on you and taking away resources.

To the guy who noticed, that he had permanent 14W package power when running MSI AB btw, that is a bug with MSI AB started with one of the last Nvidia drivers (maybe up from around 399 because I noticed it too from somewhen there around). Seems, that it always triggers the dGPU even when just having it running and no games open:

https://forums.guru3d.com/threads/m...tes-nvidia-optimus-switching-graphics.423182/

It wont be fixed it seems, because MSI AB doesnt "officially" support laptops, and dev doesnt want to invest time into the issue.

I havent tested 411.70 though yet and am still on 411.63. The user of that thread reported that from 411.70 it breaks Optimus (always triggers dGPU when running MSI AB), and I didnt want to test that out so far, which would be really bad.

Best option is to just open it when you want to game, and close when youre done. Dont make it autostart.

 

Shoot. Thanks for testing.

 

New BIOS 1.5.0 Fixes the touchscreen being disabled after screen lid closed/open fixed.

 

Thanks but have been using OOSU for years. I actually don't mind Windows update which I think usually works quite well these days, but not in this specific case. Just don't have the time to pick each update these days, especially as a .NET developers which mean LOTS of updates

Sent from my SM-N960F using Tapatalk

 

You may not have the two settings activated, I was talking about because they are not in the advised default settings list of OOSU. You have to manually activate them, and also have to activate them again after most Windows updates because Windows resets them with some updates, especially major updates.

They wont deactivate Windows updates in general, it is about driver updates via windows updates, which you were talking about. If you do it right, Windows wont install driver updates and roll over older GPU drivers, or wifi drivers, or download Dell or Logitech tools.

 

Bingo. I know they are there and I have just decided not to activate them (so far). Trying to find a mix between constant maintenance and enough stability/performance isn't to easy always. Thanks for sharing the info though, I guess many don't know about this.

 

There is no good reason for having them activated because the forced Windows drivers via Windows Updates are always months behind than the one you should use. Your problem is best example for it, but also I could tons of more. The forced Windows Update drivers for Intel 9260 card for example are from a few months ago, where the current version 20.80.0 by Intel from mid Sep 2018 has an important fix for BSOD issues.



Same for the latest Intel RST driver (Windows update version months ahead), also a BSOD fix.

And then, Windows tries to roll back older drivers over newer you had installed manually, if you havent deactivate that, which is mostly the worst of it and a stupid design by MS.

 

While that might be true, the above mentioned case is the only one I am aware of that Windows has forcefully downgraded drivers where you have installed newer yourself. This included the driver for the 9560 which is current on my machine thanks to Intels updater (which seem to be happy with other Intel drivers installed). Thats is not to say that I frequently check all drivers to see if Windows has downgraded anything, don't really have time for that.

 

Must say that after having upgraded to 1.5.0, I don't find the fans all that disturbing any more so maybe they changed the fan curve a bit. CPU is still "stuck" at around 4 Ghz even in balanced mode, but the actual temps (with stock paste) is not that bad at all.

Has there been a change in the last year or something that makes idling CPUs still report/being in higher clock although no load? Used to be that they clocked down when idle. Not using XTU or TS on this machine (yet). Used to be that high clock speed meant high temps, but now I'm idling on about 50 degrees with almost no fan activity and CPU reported as being at 4+ Ghz by Task manager and HWInfo. Must Have not had time to really follow what has changed in the last generation of Intel CPUs besided core increase.

(This is on the i9 BTW).

Edit:

Perhaps I should add that if I switch to the Power saver profile, the CPU is actually downclocked. Also does not have the Dell Power Command or whatever it is called installed, thinking that the less software by Dell installed, the better.

 

Now this could be interesting for someone looking into a performance version of the XPS15 with a numpad...

https://www.notebookcheck.net/Bette...els-and-GTX-1070-Max-Q-graphics.320842.0.html

Ugly with those "vings" though, why can't manufacturers stop this "overdesigning"? I guess the answer is that it had to look different from the current one...

 

I've never seen something that ugly before, except for Acer laptops. Not sure why the industry wants to win the prize for ugliest laptop ever year after year, for every gaming laptop they come out with. The more it costs, the uglier it gets, the heavier it gets, the louder and cheap looking. That deserves a prize if you can get the done. There are $3999 "gaming" laptops, which weight 4kg, are loud as a jet engine, are made out of plastic, the display looks like from 1996 with ugly bezels and it uses 2 power adapters with 2 cables. Even if I had the money, I would just burn it before buying something like that.

 

Well, MSI kind of invented the category ugly laptops while I agree that ASUS really gives them a run for the money these days

 

That... "thing" costs 4800Eur!!!

 

So I've been messing with the bioses for the past 2 hours but I haven't made any headway. It seems that the Dell Bios Guard prevents you from being able to extract the internal forms representation (my guess anyway). I can't seem to figure out a way around it. When you use an IFR extractor it detects the setup_vars as EFI and not UEFI (and then crashes). Sadly this is just above my skill level and I don't have any time to mess with it. If anyone wants the extracted 899407D7-99FE-43D8-9A21-79EC328CAC21 data from 1.3.0, 1.4.1, and 1.5.0 I can send it if you would like to try. Sadly I am stumped as to how to extract the form data, I had planned on running a text comparison using notepad++ to see if anything was obvious.

 

After looking at a TON of variables I'm thinking it's sadly not a EFI variable that has changed but the VBIOS. Anyone running a bios <1.4.1 mind running GPU-Z and posting your bios version?

 

@custom90gt I am on 1.3.1 (already has the 75c limit) my vBIOS ver is 86.07.4B.00.36. I was also considering the idea that when you flash the bios, you also flash the vBIOS in the XPS laptops, as the temp limit is more a vBIOS variable...
I will be upgrading soon to 1.5 to see if Dell solved the thunderbolt drain when in sleep (and test S3 sleep which is the thing I am more excited about).

 

@improwise what you are describing is not really normal. I am safely at ide at <1,5Ghz. I don't get this stick at 4ghz behavior.
It could be caused by several things, some of which you have mentioned.

I would you suggest you to:
1) Monitor the frequency using throttlestop or something else. The windows task manager is not that accurate. It does some average among the cores.
2) Check that you are on a balanced power plan and that your SST value is not 0. On AC, ThrottleStop/FIVR tab reads that I am on a value of 83.
3) Check if you have some rogue process or driver that actually keeps your processor engaged (even if you think you are idling). I would use something like Sysinternal/Process Explorer (Admin mode) instead of the task manager.

 

I still have BIOS 1.2.2. I don't game and don't normally use apps that use nVidia GPU. I also might have an older nVidia driver as well...
But I have downloaded GPU-Z and here's the screenshot. Let me know if you need it with some other settings or checks...

 

@custom90gt

You sure, it wasnt maybe already 1.2.2 doing the change? It is the only bios I think, where Dell was officially telling, they did a thermal change:

2. Enhance thermal table for better stability and performance.

Maybe the change of 1.3.1 of:

1. Mitigate coil whine noise

is a hidden meaning for, that they changed the thermal table for the GPU too. Because I have no idea, how otherwise they would reduce coil whine via a software update. Maybe they just reduced the max power or thermal of the GPU, which in their hope would reduce a coil whine.

Could also be, that the changes are in the EC (embedded controller) firmware, and not in the EFI or GPU vbios.

 

And here's GPU-Z screenshot with BIOS 1.2.2 which says the temp limit is 78C degrees.

 

Mh. even it all says the same vBios version, maybe they changed a value and the version kept the same. So a 86.07.4B.00.36 (if they just had a 86.07.4B.00.36, they just changed a thermal value for it) wont be the same on all Bios versions maybe. Could do a checksum of the files maybe extracted?

 

It started one I upgraded to 1.4(.1) and since you seem to be at 1.3.1 that might be a good explaination for not seeing the problems

Seems like any CPU load, even just 1%, causes it to go up to close to max and stay there.

 

Interesting that it has the same vbios. I'm going basically page by page and comparing 1.2.2 to 1.3.1 (since I thought 74C GPU throttling started there). I'll let you guys know what I found out. It's a total pain btw, lol.

 

The best scenario is, that people with 1.2.2 have a different 86.07.4B.00.36 vBios than people with 1.3.1 and higher. The version is made by Nvidia, and because nothing changed of the vBios code, the version is still 86.07.4B.00.36. Dell had a 86.07.4B.00.36 and modified the max temp value of it, so even it says the same vBios version, the value just changed in it.

Can you flash a dGPU vBios on the XPS with a GPU flash tools for desktops? If so, someone with 1.2.2 should extract his vBios, so we can compare it with vBios of newer.

 

I'm a bit out of the loop with these bios changes, if I can help I'm on bios 1.2.2 and this a ss of gpu-z

And this vBios