Dell XPS 15 9560 (Early 2017)

pl · 2017-03-26 09:13:37

tuxcanfly wrote:

Ah how I miss linux-nvme 4.9. Worked perfectly with my XPS 9360.
Upgrading messed up my silent boot and worse, screwed suspend/resume which was working flawlessly.

smb1234 wrote:

If anybody reading this has an idea how to fix 4.10, let me know, I'll be happy to assist with debugging.

In the 9360 thread the latest report is that suspending works again with 4.10.5, so fingers crossed. Left my machine at the office so I can't try it but let me know if anyone else does.

gc wrote:

I am still not able to use NVIDIA, I installed nvidia 378.13-4 :

378 driver does not work with the 1050. You need to revert to 375 at the moment (or just wait for it to be fixed) but it requires some manual work, and now even more so, as touched on below...

There is a mention about this in the original post in the Graphics section, but the basic idea is to get the nvidia package through ABS, change version back to 375.26, change names and dependencies manually and create a custom package and install the resulting nvidia-dkms.

argh-ch wrote:

but now I seem to have some conflict with nvidia-libgl, if i try and remove libglvnd or libgl pacman is very clear that I'm going to have a bad time. I'm unsure how to proceed.

I think the problem is this: https://www.archlinux.org/news/mesa-wit … n-testing/

It seems we are now stuck in a situation where the up-to-date driver does not work with the GPU and downgrading it does not work with the rest of the packages. I'm not sure what the full list of packages that require downgrading is... mesa, xorg-server? This starts being a bit cumbersome to manage by hand if you want to keep other things up-to-date and compiling a downgraded nvidia-dkms.

wise86 · 2017-03-26 09:47:09

Hi All,

I found a solution to the "touchpad lost sync" problem. Basically the error happen when I was using the ps2 interface to communicate with it.
If I use the I2C interface all is fine.
For do it you have to enable on the kernel:
this things: https://wiki.gentoo.org/wiki/Dell_XPS_1 … s_Touchpad
( I put them as build in) + but as module the

 PS/2 mouse

since it seems to me that if I keep it as build in, it fights a bit during the boot since it try to connect with the touchpad and it fail
after that I saw on my dmesg this line:

[    2.714679] input: DLL07BE:01 06CB:7A13 Touchpad as /devices/pci0000:00/0000:00:15.1/i2c_designware.1/i2c-7/i2c-DLL07BE:01/0018:06CB:7A13.0001/input/input11
[    2.715018] hid-multitouch 0018:06CB:7A13.0001: input,hidraw0: I2C HID v1.00 Mouse [DLL07BE:01 06CB:7A13] on i2c-DLL07BE:01

and all went smooth.

Hope it helps.

hedgepigdaniel · 2017-03-29 01:59:39

pl wrote:

In the 9360 thread the latest report is that suspending works again with 4.10.5, so fingers crossed. Left my machine at the office so I can't try it but let me know if anyone else does.

I couldn't boot with 4.10.3 and 4.10.4 but 4.10.6 works fine so far

j9ac9k · 2017-03-29 04:09:02

Can confirm suspend works for both 4.10.5 and 4.10.6.

arapat · 2017-03-29 04:23:40

Does anyone know if there is a way to let the fingerprint reader work?

Last edited by arapat (2017-04-02 15:42:22)

patrickb · 2017-03-29 05:39:24

I haven't spend time on this yet - I'm a little busy... but the 375 nvidia drivers won't compile against the 4.10 kernel and the 378 drivers still don't support the nvidia card. Also, the TB16 dock causes all sorts of lockup issues with the 4.10 kernel (works fine with the 4.9 kernel). I've downgraded for the moment. Here are the packages I built if they are any use to anyone (they have the ALLOW_ACPI_OVERRIDE compile flag set also):

https://dl.dropboxusercontent.com/u/200 … pkg.tar.xz
https://dl.dropboxusercontent.com/u/200 … pkg.tar.xz
https://dl.dropboxusercontent.com/u/200 … pkg.tar.xz

patrickb · 2017-03-29 07:09:11

hedgepigdaniel wrote:

Unfortunately the nvidia card doesn't work with TLP unless you also add pcie_port_pm=off to the kernel parameters, which I think causes increased idle power usage. It's not required to just use bbswitch to disable it though.

For what it's worth, aside from all of the issues that crept in with 4.10, the my nvidia card was working fine in conjunction with TLP. Mostly default settings EXCEPT I added the nvidia card to the blacklist. Probably similar to your solution in terms of power usage though. It's possible it might let other devices suspend (i.e. it's more targeted thatn pcie_port_pm=off); can't say I've spent any time really looking though.

/etc/default/tlp

RUNTIME_PM_BLACKLIST="01:00.0"

tillux · 2017-03-29 08:53:44

Since https://www.archlinux.org/news/mesa-wit … n-testing/ available/detected resolutions changed from 16:9 aspect ratio resolutions to 4:3 and 5:4 resolutions (it seems EDID information is now wrong, though it wasn't before... weird). Can anyone confirm that? I.e. check xrandr or Xorg.log or your display manager for a list of resolutions.

hedgepigdaniel · 2017-03-29 08:57:40

patrickb wrote:

For what it's worth, aside from all of the issues that crept in with 4.10, the my nvidia card was working fine in conjunction with TLP. Mostly default settings EXCEPT I added the nvidia card to the blacklist. Probably similar to your solution in terms of power usage though. It's possible it might let other devices suspend (i.e. it's more targeted thatn pcie_port_pm=off); can't say I've spent any time really looking though.
/etc/default/tlp
RUNTIME_PM_BLACKLIST="01:00.0"

Ah yes, that seems like a better idea

Also I take it back, 4.10.6 also often fails to boot (in my case I think something to do with starting GDM or Xorg)

pl · 2017-04-02 07:24:46

Sadly 4.10.5 did not fix suspend and others are reporting the same in the 9360 sister thread. I did get around 12 suspends in a row without problems but then faced the old scenario of resuming and hardlocking the system, with the last pre-suspend frame visible on screen. It seems it might happen less frequently than with 4.10.1 for example. Wonder what timing/race condition/sleep state/whatever it is related to.

I'm thinking installing custom linux-lts (4.9) with ACPI_REV_OVERRIDE_POSSIBLE config tweak might be the best solution for now.

Last edited by pl (2017-04-02 07:26:47)

Naveg · 2017-04-03 03:15:36

I have also been experiencing intermittent lockups on resume (4.10.3 custom with ACPI_REV_OVERRIDE_POSSIBLE), and haven't been able to identify the cause. I thought I had noticed a pattern that it tends to hardlock on the first suspend/resume after I go from battery to AC power, but that seems crazy, and I haven't had a chance to do much experimenting to confirm it yet.

hedgepigdaniel · 2017-04-03 08:55:38

For anyone who wants to test them, here are a set of package sources that make nvidia drivers, GPU switching, and NVME APST work all at once.

Kernel: linux-lts kernel 4.9.20 with ACPI_REV_OVERRIDE_POSSIBLE set and NVME APST patches applied
Nvidia driver/utils: Based on nvidia-lts but downgraded to latest nvidia long lived branch (version 375.39)

primus doesn't work for me but if I set the default bridge to virtualgl in bumblebee.conf then optirun works

simo8989 · 2017-04-03 09:00:38

Guys, any of you tried to measure the power consumption on idle with powerstat -d 2 ?
I have a custom 4.10.8 kernel with the suggested flag, bumblebee, Nvidia 375 and tlp running. Still seems I am able to turn off the GPU sometimes, at boot. At least this is what bbswitch says.
But what about your optimal power consumption ? For me seems that in the best situation possible with the above power settings (and discrete card off) I am not able to get less than 15w on idle! That is a lot. And also I forgot to mention that powertop says that all the options are on good.

What I mean is that the power saving is still shitty in any way compared to Windows. I am just wondering if is this what we will reach when the GPU issue will be fixed by the future kernel releases ? Or there is still some margin of improvement to get something better than those 15 watts on idle?

Last edited by simo8989 (2017-04-03 09:09:35)

zedsdead · 2017-04-04 03:16:44

I see similar results to sim8989. Kernel is 4.9.17 compiled with that flag and the corresponding boot parameter. at 99.9% idle connected to wifi with the backlight at 4%, I get a little more than 13 watts. I agree it would nice to get a little better battery life, but not sure what else to do about that. I'm using tlp, powertop autotune and bumblebee as well. Haven't had any problems turning off the discrete GPU but I have had issues actually getting it to work with bumblebee/optirun. I tried a few times unsuccessfully to build the nvidia drivers myself, but I think I'll try hedgepigdaniel's packages and see if that fixes it.

j9ac9k · 2017-04-05 18:40:20

simo8989 wrote:

Guys, any of you tried to measure the power consumption on idle with powerstat -d 2 ?
I have a custom 4.10.8 kernel with the suggested flag, bumblebee, Nvidia 375 and tlp running. Still seems I am able to turn off the GPU sometimes, at boot. At least this is what bbswitch says.
But what about your optimal power consumption ? For me seems that in the best situation possible with the above power settings (and discrete card off) I am not able to get less than 15w on idle! That is a lot. And also I forgot to mention that powertop says that all the options are on good.
What I mean is that the power saving is still shitty in any way compared to Windows. I am just wondering if is this what we will reach when the GPU issue will be fixed by the future kernel releases ? Or there is still some margin of improvement to get something better than those 15 watts on idle?

I am not using any kind of modified kernel (am using some bootloader kernel flags though) and when I dim the monitor, I can usually get power consumption down to just below 20W (with wifi on).

I know this isn't the newbiecorner section of the forums, but can you post a link to how to build and make use of a custom kernel with te nvidia drivers enabled (or a short guide)?

s10e · 2017-04-06 18:17:46

gtx1050 works fine with the newly released 381.09 beta driver

arapat · 2017-04-06 21:10:27

Now that 1050 works with 381.09 beta, I am struggling with this issue: https://github.com/Bumblebee-Project/Bu … issues/764.

EDIT: actually setting "acpi_rev_override=1" (with custom compiled kernel) fixed the issue. Now everything is working fine.

Last edited by arapat (2017-04-07 06:23:33)

Woutwo · 2017-04-06 21:53:19

I've just did a clean install of Arch with the latest beta drivers from nvidia, but after a boot only the cursor shows up. At least that's something.

lolzballs · 2017-04-07 02:34:22

I've tried using the patch as well as the boot flags, with bbswitch to turn the 1050 off. Everytime, bbswitch shows that it's off, but powertop still shows the GPU as using power.

dante1991 · 2017-04-07 10:46:47

is there any full guide to install arch on a 9560?

zacariaz · 2017-04-07 13:30:58

Clarification please.
As I understand it, the nvidia driver, once it's out of beta, is fixed. Is this a correct interpretation?
Then all there's left to do is apply a patch to the kernel?
Are there other of the available kernel, Zen, CK, or what not, which can be expected to work without the need for said patch?

arapat · 2017-04-07 17:20:46

dante1991 wrote:

is there any full guide to install arch on a 9560?

There is no official wiki page for 9560 yet. But hopefully all possible issues and their fixes are discussed in this post.

j9ac9k · 2017-04-07 18:49:18

dante1991 wrote:

is there any full guide to install arch on a 9560?

Should be noted that getting Arch working on the 9560 without the nvidia card is a piece of cake by following the first suggestions of the first post in this thread, and the wiki page for the XPS 15 9550.

Woutwo · 2017-04-07 19:18:07

Installing Arch on the 9560 takes about five minutes with an USB without any problems. The 1050 is another story, I (and a lot of others) are working on that right now. For now the driver seems to install pretty well, but as soon as I enable bumblebee and bbswitch-dkms-git it gives an error with "XXXX/firmware-5.bin format unsupported".. so the kernel seems to be broken by something I've done.

lolzballs · 2017-04-08 19:37:05

Woutwo wrote:

Installing Arch on the 9560 takes about five minutes with an USB without any problems. The 1050 is another story, I (and a lot of others) are working on that right now. For now the driver seems to install pretty well, but as soon as I enable bumblebee and bbswitch-dkms-git it gives an error with "XXXX/firmware-5.bin format unsupported".. so the kernel seems to be broken by something I've done.

The firmware-5.bin error is for the atheros WiFI adapter. I have it too but everything seems to work ok.

Arch Linux

#76 2017-03-26 09:13:37

Re: Dell XPS 15 9560 (Early 2017)

#77 2017-03-26 09:47:09

Re: Dell XPS 15 9560 (Early 2017)

#78 2017-03-29 01:59:39

Re: Dell XPS 15 9560 (Early 2017)

#79 2017-03-29 04:09:02

Re: Dell XPS 15 9560 (Early 2017)

#80 2017-03-29 04:23:40

Re: Dell XPS 15 9560 (Early 2017)

#81 2017-03-29 05:39:24

Re: Dell XPS 15 9560 (Early 2017)

#82 2017-03-29 07:09:11

Re: Dell XPS 15 9560 (Early 2017)

#83 2017-03-29 08:53:44

Re: Dell XPS 15 9560 (Early 2017)

#84 2017-03-29 08:57:40

Re: Dell XPS 15 9560 (Early 2017)

#85 2017-04-02 07:24:46

Re: Dell XPS 15 9560 (Early 2017)

#86 2017-04-03 03:15:36

Re: Dell XPS 15 9560 (Early 2017)

#87 2017-04-03 08:55:38

Re: Dell XPS 15 9560 (Early 2017)

#88 2017-04-03 09:00:38

Re: Dell XPS 15 9560 (Early 2017)

#89 2017-04-04 03:16:44

Re: Dell XPS 15 9560 (Early 2017)

#90 2017-04-05 18:40:20

Re: Dell XPS 15 9560 (Early 2017)

#91 2017-04-06 18:17:46

Re: Dell XPS 15 9560 (Early 2017)

#92 2017-04-06 21:10:27

Re: Dell XPS 15 9560 (Early 2017)

#93 2017-04-06 21:53:19

Re: Dell XPS 15 9560 (Early 2017)

#94 2017-04-07 02:34:22

Re: Dell XPS 15 9560 (Early 2017)

#95 2017-04-07 10:46:47

Re: Dell XPS 15 9560 (Early 2017)

#96 2017-04-07 13:30:58

Re: Dell XPS 15 9560 (Early 2017)

#97 2017-04-07 17:20:46

Re: Dell XPS 15 9560 (Early 2017)

#98 2017-04-07 18:49:18

Re: Dell XPS 15 9560 (Early 2017)

#99 2017-04-07 19:18:07

Re: Dell XPS 15 9560 (Early 2017)

#100 2017-04-08 19:37:05

Re: Dell XPS 15 9560 (Early 2017)

Board footer