You are not logged in.
Thanks for the post, I'm a bit stuck and was wondering if you could provide guidance.
When I reboot at step 8, sddm greeter fails on boot with errors including "failed to create opengl connect for format qsurfaceformat..."
Could I be missing a kennel parameter or something to that effect? What would be the best way to revert back to my Intel only graphics method?
Thanks!
EDIT: managed to recover by uninstalling the nvidia packages; and changing my kernel parameters back.
I had the same problem. However, doing steps 9 through 12 (skipping 11) (rebooting afterwards) fixes the problem. Just login directly in the console (Ctrl+Alt+F4).
Offline
I am finding that 4.10.13 runs at 15 watts idle with TLP running on defaults. I switched to linux-lts and it went to 6.6 watts idle. Which kernel are people finding works best?
Offline
I am finding that 4.10.13 runs at 15 watts idle with TLP running on defaults. I switched to linux-lts and it went to 6.6 watts idle. Which kernel are people finding works best?
I am having the same issue with the latest kernels. 15 watt as well, but I have never been able to get lower that that. Can you provide more information about that ? Thanks.
Offline
Nvidia released a new driver on the long lived branch recently which fixes the bugs with pascal GPUs and modesetting. So it now works perfectly on the 9560 with vsync. I made some AUR packages if anyone wants to try it (nvidia-llb-dkms, nvidia-utils-llb, lib32-nvidia-utils-llb). Probably not much benefit over the beta drivers if you want to use optirun, but miles better if you want to do optimus (much better performance for gaming than optirun).
Offline
In order to boot properly in Optimus mode (nvidia card off through bumblebee) I still need the parameter acpi_rev_override=1, is it the same for you?
Offline
In order to boot properly in Optimus mode (nvidia card off through bumblebee) I still need the parameter acpi_rev_override=1, is it the same for you?
Yes I think you always need that to be able to turn it on and off
When I say optimus mode, I mean using the nvidia GPU for everything (except display). https://wiki.archlinux.org/index.php/NV … ing_nvidia, https://devtalk.nvidia.com/default/topi … onization/
Offline
Ok so here is where I am at with this, the 4.10.13-1 kernel keeps me at 15 watts idle. I rolled back to 4.10.11-1 and here is my powertop with no nvidia drivers installed. I have done nothing but install the kernel and enable tlp. Is there a way to verify the Nvidia card is running? I don't have an Nvidia drivers installed but i understand that it simply draws power even without the drivers?
The battery reports a discharge rate of 7.76 W
The power consumed was 164 J
The estimated remaining time is 11 hours, 44 minutes
Summary: 1186.7 wakeups/second, 34.0 GPU ops/seconds, 0.0 VFS ops/sec and 9.0%
Usage Events/s Category Description
2.2 ms/s 568.0 Interrupt [17] idma64.1
6.7 ms/s 95.9 Process [PID 622] compton -b
5.4 ms/s 86.7 Process [PID 3139] conky -c /home
14.1 ms/s 58.4 Process [PID 6225] xfce4-terminal
445.2 µs/s 71.9 Process [PID 305] [i915/signal:0]
0.8 ms/s 49.9 Timer tick_sched_timer
1.9 ms/s 46.7 Process [PID 476] [irq/51-DLL07BE
7.4 ms/s 23.2 Process [PID 576] /usr/lib/xorg-s
87.1 µs/s 37.7 kWork intel_atomic_helper_free_
3.8 ms/s 35.1 Process [PID 872] libinput-debug-
349.4 µs/s 32.5 Interrupt [6] tasklet(softirq)
1.4 ms/s 18.3 Process [PID 618] parcellite
152.0 µs/s 10.1 Process [PID 8] [rcu_preempt]
<ESC> Exit | <TAB> / <Shift + TAB> Navigate |
Last edited by osho741 (2017-05-07 20:31:46)
Offline
osho741 wrote:I am finding that 4.10.13 runs at 15 watts idle with TLP running on defaults. I switched to linux-lts and it went to 6.6 watts idle. Which kernel are people finding works best?
I am having the same issue with the latest kernels. 15 watt as well, but I have never been able to get lower that that. Can you provide more information about that ? Thanks.
I don't know what to make of it really. I didn't have any Nvidia drivers installed and had tlp running. I don't think the dgpu is actually on even though I have not taken any measures to disable it with bumblebee. I posted my powertop in the post above with the 4.10.11-1 kernel running.
Offline
simo8989 wrote:osho741 wrote:I am finding that 4.10.13 runs at 15 watts idle with TLP running on defaults. I switched to linux-lts and it went to 6.6 watts idle. Which kernel are people finding works best?
I am having the same issue with the latest kernels. 15 watt as well, but I have never been able to get lower that that. Can you provide more information about that ? Thanks.
I don't know what to make of it really. I didn't have any Nvidia drivers installed and had tlp running. I don't think the dgpu is actually on even though I have not taken any measures to disable it with bumblebee. I posted my powertop in the post above with the 4.10.11-1 kernel running.
Really impressed from your values, I will try to roll back too. Which processor do you have? and is it 4K or 1080p ?
You can check which GPU is on, by running:
lspci -vnnn | perl -lne 'print if /^\d+\:.+(\[\S+\:\S+\])/' | grep VGA
The card with VGA Controller will be the one that you are using now.
Last edited by simo8989 (2017-05-07 21:09:23)
Offline
zcat /proc/config.gz | grep ACPI_REV_OVERRIDE
should give you the relevant line in the currently running kernel config, i.e.
CONFIG_ACPI_REV_OVERRIDE_POSSIBLE=y
cat /proc/acpi/bbswitch
should print
OFF
Also if you do
tee /proc/acpi/bbswitch <<< ON
or
tee /proc/acpi/bbswitch <<< OFF
(run tee as root) that should turn the card on or off. If you do that and find that bbswitch doesn't consistently change from ON to OFF or vice versa then its not switching properly. That's what happenned to me before I got the acpi_rev_override flag working. Also you can turn it on and see if the power consumption remains similar or increases even further.
Last edited by hedgepigdaniel (2017-05-07 22:25:06)
Offline
Should be the same processor, Kaby Lake i7. I have the 1080p screen though. It's also set pretty dim as this screen is rather bright. I will run those commands and see what it returns. I didnt make any changes to acpi. Also I have not set up bumblebee.
Last edited by osho741 (2017-05-07 22:41:19)
Offline
I made a wiki page for the 9560 (I figured there was enough stuff that is distinct from all the other XPS 15s)
https://wiki.archlinux.org/index.php/Dell_XPS_15_9560
Hopefully if we keep it up to date things will be a bit easier to follow for new users
Offline
I made a wiki page for the 9560 (I figured there was enough stuff that is distinct from all the other XPS 15s)
https://wiki.archlinux.org/index.php/Dell_XPS_15_9560
Hopefully if we keep it up to date things will be a bit easier to follow for new users
Cool !!! Thank you, really appreciated.
Offline
So I just tried building a 4.11 kernel from this commit in the arch linux pacage: https://git.archlinux.org/svntogit/pack … 619c8c234e
Somewhere along the way that I was wrong to say that the nvme patches are included in 4.10 - they are actually only included in 4.11. So I guess that means my 6-7W idle power consumption was without NVME APST, which I guess is good news...
Bad news is that my idle power consumption is up to 8-9W with this 4.11 kernel, which does have the NVME patches. I can't see anything obviously different in the powertop and I literally just installed either kernel with no other changes and did a test for a full 2 minutes with nothing else running. confuse...
I do have a niggling doubt about whether NVME APST is actually working in 4.11, because I know that some changes were made to it to avoid bugs involving particular SSDs (https://github.com/damige/linux-nvme/issues/26). I looked at the code and the only exception I can see is that a toshiba 256GB SSD does not use NVME ASPT at all, but I don't see any such exception for the 512GB model that I have (or for the samsung PM951).
Offline
Has anyone using the Precision 5520 with the Xeon E3 had luck with suspend/resume? I was never able to get it to work reliably. I'd say it works 50% of the time after first boot when X is not running and 80% of the time when X is running. I've tried the suggestions here but have been unsuccessful in diagnosing the issue so far. It's a work laptop for me so I'm debugging as I have time. I'll post back if I have any success.
IAs far as I can tell I've tried most of the setups mentioned here including linux-lts, the current 4.10 kernel and a custom kernel for NVME SSD APST support.
Offline
So I think I worked out this NVME ASPT problem. I ended up adding nvme_core.default_ps_max_latency_us=170000 to my boot parameters, because 170000 microseconds is the highest entry+exit latency for any of the ASPT states on my toshiba 512GB SSD. You can check the states (which vary between SSDs) with
nvme id-ctrl /dev/nvme0
It's only the "non-operational" states that ASPT works with, and only the ones for which the enlat+exlat <= nvme_core.default_ps_max_latency_us are used. The default value of nvme_core.default_ps_max_latency_us is 25000 which meant that for me none of them were enabled by default.
After getting this working, I measured power consumption with everything on minimum and the screen off with
xset dpms force off; sleep 10; powertop -t 120 --html=4.10.14-nvme-170000-noscreen-2.html; xset dpms force on;
a couple of runs gave me 2.8W and 3W average.
The same thing but with the screen on minimum brightness got me 5.55W and 5.92W
This is all with linux-nvme 4.10.14
Offline
Thanks a lot, @hedgepigdaniel and others. I'm now considering having this laptop with 4K panel mostly for photographing workflow use with Arch after a long, long while or alternatively go with tried and true but boring Ubuntu.
Has anybody tried running Darktable on this new XPS 15 with and without OpenCL? If so, did you find any significant speed increase with using the GPU (GTX 1050) when compared to CPU-only?
Offline
Is anyone having issues with VirtualBox and kernel panics on this machine? KVM works fine, but as soon as I try to start up the VirtualBox GUI the laptop hangs and the caps lock key starts flashing indicating a kernel panic.
I'm using a custom kernel with the AHCI override enabled, I can see that the vbox modules are loaded correctly:
[gnunn@gnunn-xps15 ~]$ sudo lsmod | grep vbox
vboxnetflt 28672 0
vboxnetadp 28672 0
vboxpci 24576 0
vboxdrv 385024 3 vboxnetadp,vboxnetflt,vboxpci
The BIOS has the two virtualization options enabled as well.
Offline
Is anyone having issues with VirtualBox and kernel panics on this machine? KVM works fine, but as soon as I try to start up the VirtualBox GUI the laptop hangs and the caps lock key starts flashing indicating a kernel panic.
After much googling I found that it is caused by the TB16 dock. Unplugging the dock and starting virtualbox works fine. See the post here where it was originally reported:
Offline
So I think I worked out this NVME ASPT problem. I ended up adding nvme_core.default_ps_max_latency_us=170000 to my boot parameters, because 170000 microseconds is the highest entry+exit latency for any of the ASPT states on my toshiba 512GB SSD. You can check the states (which vary between SSDs) with
nvme id-ctrl /dev/nvme0
It's only the "non-operational" states that ASPT works with, and only the ones for which the enlat+exlat <= nvme_core.default_ps_max_latency_us are used. The default value of nvme_core.default_ps_max_latency_us is 25000 which meant that for me none of them were enabled by default.
After getting this working, I measured power consumption with everything on minimum and the screen off with
xset dpms force off; sleep 10; powertop -t 120 --html=4.10.14-nvme-170000-noscreen-2.html; xset dpms force on;
a couple of runs gave me 2.8W and 3W average.
The same thing but with the screen on minimum brightness got me 5.55W and 5.92W
This is all with linux-nvme 4.10.14
Thank you again for your findings! This is amazing 3W on idle.
I am looking forward to have a vacation from work just to try this out . Would you mind adding this to the wiki if you have not yet done it ? It would be very useful for everyone. Thank you !!
Last edited by simo8989 (2017-05-09 22:16:29)
Offline
So I think I worked out this NVME ASPT problem. I ended up adding nvme_core.default_ps_max_latency_us=170000 to my boot parameters, because 170000 microseconds is the highest entry+exit latency for any of the ASPT states on my toshiba 512GB SSD. You can check the states (which vary between SSDs) with
nvme id-ctrl /dev/nvme0
It's only the "non-operational" states that ASPT works with, and only the ones for which the enlat+exlat <= nvme_core.default_ps_max_latency_us are used. The default value of nvme_core.default_ps_max_latency_us is 25000 which meant that for me none of them were enabled by default.
After getting this working, I measured power consumption with everything on minimum and the screen off with
xset dpms force off; sleep 10; powertop -t 120 --html=4.10.14-nvme-170000-noscreen-2.html; xset dpms force on;
a couple of runs gave me 2.8W and 3W average.
The same thing but with the screen on minimum brightness got me 5.55W and 5.92W
This is all with linux-nvme 4.10.14
This will probably be a dumb question but may I ask how you edited the default value in the boot parameter?
Offline
This will probably be a dumb question but may I ask how you edited the default value in the boot parameter?
Offline
@hedgepigdaniel Which bios version are you on ? I am on 1.2.4 the latest. I am having a lot of crazy problems with any linux distro, the computer overheats a lot reaching peaks of 80 C with the fans spinning like crazy, the power consumption of course is extremely high. The thing is that bbswitch outputs OFF and bumblebee 0.8 correctly detects and turns off the GPU, tlp is up and running with the Nvidia exception and default configs.
Last edited by simo8989 (2017-05-10 22:45:16)
Offline
1.2.4 same as you.
Sounds like you have a process using lots of CPU?
Offline
1.2.4 same as you.
Sounds like you have a process using lots of CPU?
Nha.. that's not the problem, at least from htop seems all right. The thing is that I am trying on Fedora right now with kernel 4.10.14 and then I compile the kernel with the correct flag on yes but the situation does not change when then I boot with acpi_rev_override=1. The laptop almost melts before and after that and also the GPU is OFF, in the first case with original kernel sometimes and with custom kernel always but everything is hot and the fans spin constantly, I had those problems as peaks in temperature (80 C) actually for some minutes and not for all the time. I hope it's a Fedora problem and I will soon try with arch, btw which DE are you using? Gnome on Wayland?
I had Ubuntu 17.04 before this test and that kernel has the flag on y by default. With that Distro I had good results in power cosumption but never less than 14.5 W from powertop analysis. I am starting to think that the problem it's my computer.. it's very wired beacause on windows runs like a charm, with almost 9 h of battery life and the fans never spinning, apart when I charge the laptop.
Last edited by simo8989 (2017-05-11 07:12:30)
Offline