Problem with nvidia-persistenced not working

Question

I'm having a problem which I think I've narrowed down to the nvidia persistence daemon.

Issue: I have my HTPC connected to my Onkyo receiver. When I turn off the receiver and turn it back on, I have no signal from the HTPC. I can press Ctrl + Alt + F1 and run sudo stop lightdm then sudo start lightdm. Once I do this I'm back to the regular screen. This is not a long-term solution though as

1. it's a pain
2. I will have other users who want to watch TV and will not know what to do.

Through my googlefu I found that I need to set persistence-mode to on, though this will be going away. I have tried this method with sudo "nvidia-smi -pm 1" and it will set it to on, but it doesn't appear to work as the issue will re-occur.

I have also tried running sudo nvidia-persistenced --user [user] --persistence-mode and I can see a process start. When I turn my receiver off though, the same thing occurs.

The syslog has the following:

... nvidia-persistenced: Failed to open PID file: File exists
... nvidia-persistenced: Shutdown (5232)

I have made sure the user I have nvidia-persistenced running as is part of the nvidia-persistenced group. I have also set permissions on the /var/run/nvidia-persistencd/ folder to 777.

I'm going crazy trying to resolve this. I want to be able to turn my receiver off and when I turn it back on have the screen where I left off.

• Running Ubuntu 14.04 (Mythbuntu install)
• Video card is Nvidia GT 730 using driver 352.63

Also, once I get the nvidia-persistenced daemon working, I need to set it up to start at system startup. How would I go about doing that?

Any help is greatly appreciated. Thanks in advance.

Answer 1

I was able to resolve this issue with assistance from another forum, but here is the resolution in-case anyone is interested.

Basically follow the instructions in this link: https://www.mythtv.org/wiki/Nvidia-cards_and_no_picture_when_box_is_on_before_the_TV

In case that link is removed, here's the gist of what I did:

• Run 'nvidia-settings' on the local X-Server
• Click the entry corresponding to your TV (e.g. DFP-0) Click 'Acquire
• EDID' and save the 'edid.bin' file someplace safe, for instance
  '/etc/X11/tv-edid.bin'
• Open your existing config file, usually in /etc/X11/xorg.conf
• Find the 'Device' section for your video card.
• Add a line to force the stored EDID:
• Option "CustomEDID" "DFP-0:/etc/X11/edid.bin"
• Reboot

I ended up adding the following lines as well, not sure if they are needed or not though:

Option "ModeValidation" "NoDFPNativeResolutionCheck"
Option "ExactModeTimingsDVI" "True"
Option "UseDisplayDevice" "DFP-0"
Option "ConnectedMonitor" "DFP-0" `

Now I'm able to turn off my receiver and/or change the inputs and I can return to my HTPC exactly where I left off.

Answer 2

Try to install libcupti, it helped me.

sudo apt-get install libcupti-dev

I dont know why this is helping. But this one of the libs that Tensorflow asks for during installation. Specifically, it is listed under "NVIDIA requirements to run TensorFlow with GPU support":

The libcupti-dev library, which is the NVIDIA CUDA Profile Tools Interface. This library provides advanced profiling support.