[Plugin] Nvidia-Driver

[ich777]

1 hour ago, kokonator said:

Sorry, I'm very new to all of this.

No worries, please post your Diagnostics as @alturismo already wrote.

[kokonator]

20 hours ago, ich777 said:

No worries, please post your Diagnostics as @alturismo already wrote.

I think I figured it out, turns out I just had to remap my media libraries.  I don't think it was an issue with the plug-in after all.

[bombz]

Hello, 

Reaching out as I am having concerns with Plex using the GPU for hw transcode playback.

I have Jellyfin also installed and transcoding works with the GPU, is there something I am missing ?

Extra Parameters:

--runtime=nvidia

 

NVIDIA DC:
 

Container Variable: NVIDIA_DRIVER_CAPABILITIES

all

 

NVIDIA VD:

Container Variable: NVIDIA_VISABLE_DEVICES

GPU-ID HERE

 

Not sure why, this seems to work with all other dockers expect for plex
Any feedback would be apprecaited.

Edited December 9, 2023 by bombz

[alturismo]

28 minutes ago, bombz said:

Any feedback would be apprecaited.

may try with the official container in terms you have a 3rd party like binhex, linuxserver, ... running.

[bombz]

2 hours ago, alturismo said:

may try with the official container in terms you have a 3rd party like binhex, linuxserver, ... running.

Using officiial
plexinc/pms-docker

[alturismo]

7 minutes ago, bombz said:

Using officiial

well, then may post your diagnostics and also your plex docker run command, should work ...

[ich777]

1 hour ago, bombz said:

Using officiial
plexinc/pms-docker

Please try reinstalling it and specify a new token.

Are you able to issue nvidia-smi in the container?

[bombz]

here is SMI

 

Here is the docker run

docker run
  -d
  --name='Plex-Media-Server'
  --net='eth1'
  --ip='192.168.1.2'
  -e TZ="America/New_York"
  -e HOST_OS="Unraid"
  -e HOST_HOSTNAME="UnRAID"
  -e HOST_CONTAINERNAME="Plex-Media-Server"
  -e 'PLEX_CLAIM'='claim-TOKEN HERE
  -e 'PLEX_UID'='99'
  -e 'PLEX_GID'='100'
  -e 'VERSION'='latest'
  -e 'NVIDIA_DRIVER_CAPABILITIES'='all'
  -e 'NVIDIA_VISABLE_DEVICES'='GPU-867b6976-8663-b63c-234e-59c42267ad88'
  -l net.unraid.docker.managed=dockerman
  -l net.unraid.docker.webui='http://[IP]:[PORT:32400]/web'
  -l net.unraid.docker.icon='https://raw.githubusercontent.com/plexinc/pms-docker/master/img/plex-server.png'
  -v '/tmp':'/transcode':'rw'
  -v '/mnt/':'/data':'rw'
  -v '/mnt/user/CACHE_PLEX/appdata/Plex-Media-Server/':'/config':'rw'
  --runtime=nvidia 'plexinc/pms-docker'

8bba84b89978daba07238055496e246f35aa4eb244db997964e52789f7126453

The command finished successfully!

 

Maybe a full server reboot would kick it into gear?

 

[bombz]

Resolved over an embarrassing concern... typo in the variable -- unreal @ me !

[ich777]

7 hours ago, bombz said:

VISABLE

Sorry haven‘t got time yet to reply.

 

I see it, anyways, glad that it‘s solved!

[thewave]

I recently upgraded from a Quadro P4000 to a Quadro RTX A4000 and I'm noticing some problematic behavior.. It looks like dockers that rely on the GPU do not function correctly.  Looking at syslog, I see this error a few times per minute:

 

kernel: nvidia-modeset: WARNING: GPU:0: Unable to read EDID for display device DP-0

 

Notably, if I plug the server into a monitor and reboot everything appears to function correctly.  I don't have a dummy displayport plug on hand, but I suspect that may improve the situation.  Regardless, this wasn't an issue with the P4000 and the server was perfectly happy functioning headless.

 

Does anyone have any ideas about how I could fix this?

[alturismo]

18 minutes ago, thewave said:

Does anyone have any ideas about how I could fix this?

get a dummy plugin ...

 

Quadro cards are usually server cards and dont rely on a monitor, RTX are main stream cards and like to rely on a monitor ... a simpl edummy plug will solve it (as you noticed yourself already) as mentioned several times here before

[thewave]

Just now, alturismo said:

get a dummy plugin ...

 

Quadro cards are usually server cards and dont rely on a monitor, RTX are main stream cards and like to rely on a monitor ... a simpl edummy plug will solve it (as you noticed yourself already) as mentioned several times here before

Thanks for the reply - you'll note that I mentioned I upgraded from one Quadro to a different Quadro that is two generations newer; which is partly why I didn't expect a dummy plug to be necessary

[alturismo]

8 minutes ago, thewave said:

Thanks for the reply - you'll note that I mentioned I upgraded from one Quadro to a different Quadro that is two generations newer; which is partly why I didn't expect a dummy plug to be necessary

yep, i saw that, but in the end as you see in the logs ... you actually do  sorry to say ...

 

but i agree, i wouldnt have expected it either ...

[Maginos]

Hi guys, 

 

I have a question concerning the GT 710 in my unraid server: 

How can I tell unraid, that it should use the GT 710 for the output of the console instead of the iGPU from my i5-11400?

 

The background of this is, that I want to use the iGPU for plex or handbrake and the GT 710 for my pikvm. I tried different settings in the BIOS concerning primary display, but everything else except for "Auto" didn't give any output on the corresponding port (available options are "CPU GPU" and "PCIE"). 

What I can observe is, that when I connect the pikvm to the HDMI port of the GT 710, the console output stops at a certain point in the boot process, while at the HDMI port of the MB, it goes directly to the login shell.

 

This has worked without any issues in the past before, but for a few months, it doesn't work any longer.

Here is my syslinux configuration:
 

kernel /bzimage
append pcie_acs_override=downstream,multifunction vfio_iommu_type1.allow_unsafe_interrupts=1 initrd=/bzroot

 

Has anyone an idea, what I can do about this?

Thank you for your help!

[ich777]

2 minutes ago, Maginos said:

How can I tell unraid, that it should use the GT 710 for the output of the console instead of the iGPU from my i5-11400?

Usually you have to set this in the BIOS which GPU is the preferred one and then replug the cable from your monitor to the correct one after changing the setting in the BIOS.

 

3 minutes ago, Maginos said:

The background of this is, that I want to use the iGPU for plex or handbrake and the GT 710 for my pikvm.

You don't need a second GPU for this scenario. It is possible to use the iGPU for Plex and for the display output/PiKVM.

 

...and if you are worried about giving up performance when doing it like that for transcoding for Plex, its negligible.

[Maginos]

1 minute ago, ich777 said:

Usually you have to set this in the BIOS which GPU is the preferred one and then replug the cable from your monitor to the correct one after changing the setting in the BIOS.

 

I tried this. I had connected the pikvm to the GT 710 and changed then the BIOS setting to "PCIE". Then there was no console output after the screen with the blue background and the different boot options for unraid. So I switched back to "Auto".

 

3 minutes ago, ich777 said:

You don't need a second GPU for this scenario. It is possible to use the iGPU for Plex and for the display output/PiKVM.

 

...and if you are worried about giving up performance when doing it like that for transcoding for Plex, its negligible.

Thank you, I didn't know about that! Again what learnt. 

[ich777]

4 minutes ago, Maginos said:

I tried this. I had connected the pikvm to the GT 710 and changed then the BIOS setting to "PCIE". Then there was no console output after the screen with the blue background and the different boot options for unraid. So I switched back to "Auto".

Hmmm... this is strange, did you change anything or did a BIOS update?

 

4 minutes ago, Maginos said:

Thank you, I didn't know about that! Again what learnt. 

So it's now "solved"?

[Maginos]

5 minutes ago, ich777 said:

Hmmm... this is strange, did you change anything or did a BIOS update?

Nope.

 

5 minutes ago, ich777 said:

Hmmm... this is strange, did you change anything or did a BIOS update?

 

So it's now "solved"?

It seems so, yes.

The pikvm is connected to the HDMI port of the MB and the CLI of unraid is responsive. At the same time, handbrake is starting with passed through /dev/dri device and is converting a mov file to m4v with my custom QSV preset without any issue. So I guess my problem is solved.

Thank you ich777 for your fast response and your help!

[VolzanIT]

good morning, my gpu after a few seconds stops being recognized by nvidia drivers and nvidia-smi, although the video output works properly

the card is attached to pcie i try to chenge pcie slot too but without any success

 

*-display                 
       description: VGA compatible controller
       product: TU117 [GeForce GTX 1650]
       vendor: NVIDIA Corporation
       physical id: 0
       bus info: pci@0000:01:00.0
       version: a1
       width: 64 bits
       clock: 33MHz
       capabilities: pm msi pciexpress vga_controller cap_list rom
       configuration: driver=nvidia latency=0
       resources: irq:16 memory:de000000-deffffff memory:c0000000-cfffffff memory:d0000000-d1ffffff ioport:e000(size=128) memory:c0000-dffff

 

edit in the file i noticed this

1

Dec 14 13:38:48 nas kernel: NVRM: Xid (PCI:0000:01:00): 79, pid='<unknown>', name=<unknown>, GPU has fallen off the bus.
Dec 14 13:38:48 nas kernel: NVRM: GPU 0000:01:00.0: GPU has fallen off the bus.

nas-diagnostics-20231214-1354.zip nvidia-bug-report.log.gz

Edited December 14, 2023 by VolzanIT

[ich777]

13 minutes ago, VolzanIT said:

good morning

Please upgrade your Unraid version to 6.12.6 and if the issue persists please upload new Diagnostics after the error occurred from the latest Unraid version.

 

13 minutes ago, VolzanIT said:

nvidia-bug-report.log.gz

You don't need to upload that since everything that I need is included in the Diagnostics.

[VolzanIT]

1 hour ago, ich777 said:

Please upgrade your Unraid version to 6.12.6 and if the issue persists please upload new Diagnostics after the error occurred from the latest Unraid version.

 

You don't need to upload that since everything that I need is included in the Diagnostics.

this is the new diagnostic after the update

nas-diagnostics-20231214-1515.zip

[ich777]

15 minutes ago, VolzanIT said:

this is the new diagnostic after the update

As you can see here you have a Code 79 XID error (like it was before):

NVRM: Xid (PCI:0000:01:00): 79, pid='<unknown>', name=<unknown>, GPU has fallen off the bus.

(You can read more about that here in the table of possible causes)

 

Sadly enough this error is a bit hard to troubleshoot since it can mean almost anything like: that you have a faulty card (frame buffer error), aerror in the PCI bus, issues with the power supply, a thermal issue or an issue with the driver (the last one is actually unlikely because on other systems the driver is working as expected since the plugin always uses the same driver package which is pre-compiled).

 

Did you change anything in your system, a BIOS setting, added hardware or similar,...?

 

Are you also sure that your power supply is still up to the task and can deliver enough juice to run the card properly (IIRC the last user had a faulty power supply)?

 

Please also make sure that you enable Above 4G Decoding in your BIOS and Resizable Bar Support (if you have that option) in your BIOS.

Have you yet tried to boot with UEFI instead of Legacy mode (don't forget to allow UEFI boot)?

 

You could also try, if this is possible for you, to put the card in another system, install the driver and put some 3D load on it for half an hour and see if the card is working properly.

[VolzanIT]

ok ill try with another card and put this one on the other system i have

[VolzanIT]

2 hours ago, ich777 said:

As you can see here you have a Code 79 XID error (like it was before):

NVRM: Xid (PCI:0000:01:00): 79, pid='<unknown>', name=<unknown>, GPU has fallen off the bus.

(You can read more about that here in the table of possible causes)

 

Sadly enough this error is a bit hard to troubleshoot since it can mean almost anything like: that you have a faulty card (frame buffer error), aerror in the PCI bus, issues with the power supply, a thermal issue or an issue with the driver (the last one is actually unlikely because on other systems the driver is working as expected since the plugin always uses the same driver package which is pre-compiled).

 

Did you change anything in your system, a BIOS setting, added hardware or similar,...?

 

Are you also sure that your power supply is still up to the task and can deliver enough juice to run the card properly (IIRC the last user had a faulty power supply)?

 

Please also make sure that you enable Above 4G Decoding in your BIOS and Resizable Bar Support (if you have that option) in your BIOS.

Have you yet tried to boot with UEFI instead of Legacy mode (don't forget to allow UEFI boot)?

 

You could also try, if this is possible for you, to put the card in another system, install the driver and put some 3D load on it for half an hour and see if the card is working properly.

ok tryed with a gt710 () it seams to not work i get this error NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.

Edit NVM Need legacy driver for this gpu

 

Edited December 14, 2023 by VolzanIT