Ubuntu 16.04 - Nvidia 367.57 - GTX 1060 - Bad Performance / Video Output Crash after resume from suspend

I have several issues with my new GTX 1060, which result in video output crash. The monitor output is stuck, displaying the same output all the time. This happens when resuming from suspend to ram.

The crash occurs after inserting correct password credentials at login screen.
(Moving the cursor and inserting the password with visual feedback is fine on login screen after resume.)

With the following steps I can reproduce the issue:

  • Install Ubuntu 16.04 (UEFI mode)
  • apt-get update
  • apt-get dist-upgrade
  • apt-get autoremove --purge xserver-xorg-video-*
  • apt-get install nvidia-367 openssh-server*
  • reboot
  • systemctl suspend
  • Resume from suspend

Sometimes the X-Server Session can be rescued by switching to a login shell (ctrl+alt+f6) and switching back to X. In other cases both, X-Session and login shell are not available anymore (login shell prompt not shown, just black screen.)

test@ubuntu:~$ uname -r
4.4.0-59-generic

test@ubuntu:~$ lspci -v | perl -ne '/VGA/../^$/ and /VGA|Kern/ and print' 
01:00.0 VGA compatible controller: NVIDIA Corporation Device 1c03 (rev a1) (prog-if 00 [VGA controller])
	Kernel driver in use: nvidia
	Kernel modules: nvidiafb, nouveau, nvidia_367, nvidia_367_drm

test@ubuntu:~$ dmesg | grep NVRM | tail -5
[  119.691681] NVRM: Xid (PCI:0000:01:00): 56, CMDre 00000001 00000094 000100a3 00000007 00000000
[  119.691696] NVRM: Xid (PCI:0000:01:00): 56, CMDre 00000002 00000094 000100a4 00000007 00000000
[  119.691900] NVRM: Xid (PCI:0000:01:00): 56, CMDre 00000001 000000c0 000100b2 00000007 00000000
[  119.691910] NVRM: Xid (PCI:0000:01:00): 56, CMDre 00000002 000000c0 000100b3 00000007 00000000
[  120.183980] NVRM: Xid (PCI:0000:01:00): 56, CMDre 00000000 00000080 00000000 00000005 00000021

Nvidia Bug Report: https://www.dropbox.com/s/qfnrakn85gr27h9/nvidia-bug-report-2017-01-11.log.gz?dl=1

Update

Using just one monitor connected via DVI (in the above mentioned issue two monitors are connected) and using the kernel parameters “rdblacklist=nouveau nouveau.modeset=0 vga=0” the same issue occurs.

As confirmed by three Windows-using GTX 1060 owners thus far, there’s a seeming co-relation between the presence of the nVidia driver’s audio component and resume-from-suspend failures in the GTX 1060 (starting at Winterflareon’s post #53):

GTX 1060 Windows sleep S3 problem - GeForce Forums
https://forums.geforce.com/default/topic/955352/geforce-1000-series/gtx-1060-windows-sleep-s3-problem/4/

Thanks for the reply.

Actually the issue for me started in ArchLinux, where I get the same error code Xid 56, during ‘normal desktop usage’.

The issue starts that a window of one applications starts to flicker and after some time the monitors turn black. The video output of the X-Session somtimes can be rescued (as described above) by switching to login shell and back. (Sometimes this may rescue the X-Session, in other cases the login terminal is just black, too).

Or I have a similiar issue as described in https://devtalk.nvidia.com/default/topic/979687/linux/arch-linux-on-nvidia-375-20-3-driver-mouse-disappears-virtual-consoles-blank-/.

I tried to narrow it down to a more ‘simple case’ and wanted to avoid that there might be some configuration issue in my ArchLinux setup. I ended with Ubuntu and suspend to ram…

Is there a way of disabling the nVidia driver’s audio function under ArchLinux?

Just to clarify, resume-from-suspend works under Ubuntu (ver.?, kernel?, system specs?).

Yesterday i was able to reproduce the issue on Ubuntu (resume after suspend) several times…

This morning I it just occured once (suspending and resuming about 20 times).

But as mentioned above, the issue/freeze occurs after inserting the password at lock screen.
(As shown in time difference of dmesg.)

[   28.096150] r8169 0000:03:00.0 enp3s0: link down
[   28.096210] IPv6: ADDRCONF(NETDEV_UP): enp3s0: link is not ready
[   33.526898] r8169 0000:03:00.0 enp3s0: link up
[   33.526905] IPv6: ADDRCONF(NETDEV_CHANGE): enp3s0: link becomes ready
[   83.973402] NVRM: GPU at PCI:0000:01:00: GPU-abd4bbe5-adbe-36dd-de63-18ce9de7a520
[   83.973412] NVRM: GPU Board Serial Number: 
[   83.973426] NVRM: Xid (PCI:0000:01:00): 56, CMDre 00000001 00000094 000100a3 00000007 00000000
[   83.973559] NVRM: Xid (PCI:0000:01:00): 56, CMDre 00000001 000000c0 000100a8 00000007 00000000
[   83.995777] NVRM: Xid (PCI:0000:01:00): 56, CMDre 00000001 00000094 000100a3 00000007 00000000
[   83.995892] NVRM: Xid (PCI:0000:01:00): 56, CMDre 00000001 000000c0 000100b2 00000007 00000000

https://www.dropbox.com/s/f3hi6so4uvgv3uc/dmesg-2017-01-11.log.gz?dl=1

Yes, resume from suspend works under Ubuntu 16.04 (but not always):

  • NVIDIA 367.57
  • Kernel: Linux ubuntu 4.4.0-59-generic
  • System Specs:
    • Gigabyte GA-H87-HD3
    • Intel i5-4570
    • DDR3 16GB - Corsair Vengeance 4x4GB - CML8GX3M2A1600C9
    • PSU: be quiet! Straight Power E9

If more detailed specifications are required, i can provide them.

I can blacklist the kernel module snd_hda_intel, which disables Nvidia HDMI audio, but also onboard audio…
I am not sure if it is possible to disable Nvidia audio only.

Using only one DVI connected monitor seems to have the lowest failure rate.
I have been using a DP to HDMI cable for my second monitor which resulted in higher failure rate.
But using default the HDMI port and default HDMI cable also results in high failure rate.

I have blacklisted snd_hda_intel and it was not any better…

Wanted to drop this in here: https://devtalk.nvidia.com/default/topic/962231/linux/resume-from-suspend-freezes-system-gtx-970-arch-linux-kernel-4-4-4-7-nvidia-370-/

Arch had been having the same issue, but the NVIDIA rep states that it’s fixed in the next release.