[Plugin] Linuxserver.io - Unraid Nvidia


Recommended Posts

Just read the work that went into releasing the latest version. Thank you to all that are involved in keeping this amazing addition to unraid working. I've been enjoying the benefits of it ever since it was released and as we all know how great it is to have this ability. I really appreciate it.

I searched back a bit and didn't see anything recently about the work being done on combining the Nvidia build and the DVB build. That would be the next dream come true. Is there any place I can follow the development of that build? Seems like it's getting lost in this thread.
Yeah, I'm just too tied up with work at the moment, I'll try and pick that up in the New Year

Sent from my Mi A1 using Tapatalk

Link to comment

Hi guys,

 

I've been using Unraid Nvidia for a couple months now with Plex and its been working great.  I did notice however that whenever the Plex docker was restarted the fan on my P2000 would ramp up to about 90% until Plex was restarted.  This wasn't bothersome as I always kept Plex running, however when I recently switched to Emby I noticed that the GPU fan would ramp to 75-90% whenever there were no running processes in nvidia smi, so as soon as a transcode completes the fan speeds up.  I have tried to find solutions to this, but haven't come across anyone speaking of the same issue.  My server is on my desk and the fan is tolerable but not ideal, does anyone know of a way to fix this?  I am running version 6.7.2 and the fan is usually at about 55% with the GPU at 55-60C with a transcode running.  At idle it is usually about 80% with the GPU at 35-40C.

 

Thank you all.

Link to comment
3 hours ago, satmeiler said:

Hi guys,

 

I've been using Unraid Nvidia for a couple months now with Plex and its been working great.  I did notice however that whenever the Plex docker was restarted the fan on my P2000 would ramp up to about 90% until Plex was restarted.  This wasn't bothersome as I always kept Plex running, however when I recently switched to Emby I noticed that the GPU fan would ramp to 75-90% whenever there were no running processes in nvidia smi, so as soon as a transcode completes the fan speeds up.  I have tried to find solutions to this, but haven't come across anyone speaking of the same issue.  My server is on my desk and the fan is tolerable but not ideal, does anyone know of a way to fix this?  I am running version 6.7.2 and the fan is usually at about 55% with the GPU at 55-60C with a transcode running.  At idle it is usually about 80% with the GPU at 35-40C.

 

Thank you all.

I can't remember the source, but I added the below script using the Userscripts plugin and have it running hourly (need to update to a shorter time window) that will check to see if the gpu is being used, otherwise return it to an idle state of P0. Others had cited that once the card was fired up it would not return to idle once done, this script check for any processes and then forces the return. See if it helps for you.

 

#!/bin/bash
TOKILL=""
CPT_TRANSCODE=-1
CPT_SERVER=-1

for ELEMENT in $(lsof /dev/nvidia0 | awk '{print $1 "@" $2}')
do
 IFS='@' read -r -a array <<< "$ELEMENT"
 if [ "${array[0]}" != "Xorg" ] && [ "${array[0]}" != "ffmpeg" ]
 then
  if [ "${array[0]}" = "Plex\x20T" ]
  then
  ((CPT_TRANSCODE++))
  fi
  if [ "${array[0]}" = "Plex\x20M" ]
  then
   ((CPT_SERVER++))
   TOKILL="${array[1]}"
  fi
 fi
done

if [ $CPT_SERVER -gt 0 ] && [ $CPT_TRANSCODE -lt 1 ] && [ "$1" != "test" ]
then
 kill -9 "$TOKILL"
fi
if [ "$1" == "test" ]
then
 echo "$TOKILL"
fi

 

Edited by dewlite
  • Thanks 1
Link to comment
5 hours ago, dewlite said:

I can't remember the source, but I added the below script using the Userscripts plugin and have it running hourly (need to update to a shorter time window) that will check to see if the gpu is being used, otherwise return it to an idle state of P0. Others had cited that once the card was fired up it would not return to idle once done, this script check for any processes and then forces the return. See if it helps for you.

 


#!/bin/bash
TOKILL=""
CPT_TRANSCODE=-1
CPT_SERVER=-1

for ELEMENT in $(lsof /dev/nvidia0 | awk '{print $1 "@" $2}')
do
 IFS='@' read -r -a array <<< "$ELEMENT"
 if [ "${array[0]}" != "Xorg" ] && [ "${array[0]}" != "ffmpeg" ]
 then
  if [ "${array[0]}" = "Plex\x20T" ]
  then
  ((CPT_TRANSCODE++))
  fi
  if [ "${array[0]}" = "Plex\x20M" ]
  then
   ((CPT_SERVER++))
   TOKILL="${array[1]}"
  fi
 fi
done

if [ $CPT_SERVER -gt 0 ] && [ $CPT_TRANSCODE -lt 1 ] && [ "$1" != "test" ]
then
 kill -9 "$TOKILL"
fi
if [ "$1" == "test" ]
then
 echo "$TOKILL"
fi

 

Thank you, Ill give that a shot later today when I have time, I don't have any experience installing a script into the Nvidia Drivers, but I assume there is documentation about that.  Ill report back if that fixes it.

Link to comment
20 hours ago, dewlite said:

I can't remember the source, but I added the below script using the Userscripts plugin and have it running hourly (need to update to a shorter time window) that will check to see if the gpu is being used, otherwise return it to an idle state of P0. Others had cited that once the card was fired up it would not return to idle once done, this script check for any processes and then forces the return. See if it helps for you.

 


#!/bin/bash
TOKILL=""
CPT_TRANSCODE=-1
CPT_SERVER=-1

for ELEMENT in $(lsof /dev/nvidia0 | awk '{print $1 "@" $2}')
do
 IFS='@' read -r -a array <<< "$ELEMENT"
 if [ "${array[0]}" != "Xorg" ] && [ "${array[0]}" != "ffmpeg" ]
 then
  if [ "${array[0]}" = "Plex\x20T" ]
  then
  ((CPT_TRANSCODE++))
  fi
  if [ "${array[0]}" = "Plex\x20M" ]
  then
   ((CPT_SERVER++))
   TOKILL="${array[1]}"
  fi
 fi
done

if [ $CPT_SERVER -gt 0 ] && [ $CPT_TRANSCODE -lt 1 ] && [ "$1" != "test" ]
then
 kill -9 "$TOKILL"
fi
if [ "$1" == "test" ]
then
 echo "$TOKILL"
fi

 

I installed the user scripts plugin and ran that script however I saw now change in the GPU temp.  When I clicked view log, the log showed that the actual log was stored in the /tmp folder.  I attempted to use midnight commander to copy that txt file somewhere it could be accessed.  Midnight commander seems to work at random for me I'm not sure if I'm missing something but I was unable to perform any actions of the file.  To install the script I mounted the flash share and browsed to the file created when I added a script with the plugin.  I then copied the script from this post into that txt file and ran it.  Am I doing all of this right?

Link to comment
4 hours ago, satmeiler said:

hen I clicked view log, the log showed that the actual log was stored in the /tmp folder.  I attempted to use midnight commander to copy that txt file somewhere it could be accessed.

Why not simply hit the download log button?

Link to comment

Let me start off by saying, thank you for the work that goes into this plugin. Honestly should be something that is built into Unraid by default. Anyway, I am now upgraded to 6.8 RC9 but I do not see my quadro card as an availble card.

image.png.3cdf444b626a729f15547deaaac9dbee.png

 

But using watch nvidia-smi I see my quadro (1080 is being used in a VM)

 

image.png.0c80b23740bb2a1630574b27c58f8792.png

 

Any reason why I no longer see the P2000?

 

Thank you

Link to comment
3 hours ago, Eggman1414 said:

Let me start off by saying, thank you for the work that goes into this plugin. Honestly should be something that is built into Unraid by default. Anyway, I am now upgraded to 6.8 RC9 but I do not see my quadro card as an availble card.

image.png.3cdf444b626a729f15547deaaac9dbee.png

 

But using watch nvidia-smi I see my quadro (1080 is being used in a VM)

 

image.png.0c80b23740bb2a1630574b27c58f8792.png

 

Any reason why I no longer see the P2000?

 

Thank you

Probably since it's passed through to your vm?

Link to comment
Let me start off by saying, thank you for the work that goes into this plugin. Honestly should be something that is built into Unraid by default. Anyway, I am now upgraded to 6.8 RC9 but I do not see my quadro card as an availble card.
image.png.3cdf444b626a729f15547deaaac9dbee.png
 
But using watch nvidia-smi I see my quadro (1080 is being used in a VM)
 
image.png.0c80b23740bb2a1630574b27c58f8792.png
 
Any reason why I no longer see the P2000?
 
Thank you
Is the Quadro still supported on the driver version?

Sent from my Mi A1 using Tapatalk

Link to comment
  • trurl locked this topic
Guest
This topic is now closed to further replies.