9 minutes ago, Nanobug said:

Looks fine to me, and it's transcoding as well:

Please make a short post on the GPU Statistics support thread with the screenshot from the output from nvidia-smi

47 minutes ago, ich777 said:

Please make a short post on the GPU Statistics support thread with the screenshot from the output from nvidia-smi

I will, thank you

I have noticed recently that GPU transcoding isn't work in Plex. I also tried GPU transcoding in Tdarr and Tdarr was using a lot of CPU. I have been using GPU on Plex for months now with no issues. I am on Unraid 6.9.2 with Nvidia driver 460.80. I am not sure when GPU transcoding stopped working. 

 

I tried removed the variables from the template and adding them. I tried rebooting Unraid and downloading new versions. I cannot seem to get Plex or any container to use GPU.

Hi, not entirely sure if this is the right place to post this but here goes:

 

My setup:
    - CPU: R9 3900X
    - Motherboard: Asus Crosshair VIII Hero
    - PCIe x16 Top Slot: GTX 1080 Ti
    - PCIe x16 Second Slot: Quadro P2000
    - PCIe x16 Third Slot: LSI 9207-8i
    - Running unRAID 6.9.2

 

What I want to accomplish:
    - Pass through the primary GPU (1080Ti) for a W10 VM for gaming
    - Use the secondary GPU (P2000) for Plex/Emby hardware transcoding

 

From what I understand, I need to:
    1) Dump the vBIOS (following SpaceInvaderOne's video) for the 1080Ti since it's an nVidia GPU in the primary slot
    2) Install this nVidia plugin to use the P2000 for hardware transcoding in Docker

 

My question is:

Apart from 1 & 2 above, is there anything special I need to do to accomplish my goals (e.g. stubbing the primary GPU or something like that)?

 

N.B.: if switching the GPUs (i.e. put the P2000 in the primary slot) would somehow make things easier, unfortunately I can't. My 1080Ti is a 2.5 slot card and there isn't enough clearance between the second PCIe slot and the LSI HBA in the third PCIe slot.

3 hours ago, exdox77 said:

I tried removed the variables from the template and adding them. I tried rebooting Unraid and downloading new versions. I cannot seem to get Plex or any container to use GPU.

Please open up a console and can you give me the output from 'nvidia-smi'?

 

Can you also send me a secreenshot from the container template itself (with Advanced View turned on)?

 

9 minutes ago, rragu said:

Pass through the primary GPU (1080Ti) for a W10 VM for gaming

Go to Tools -> System Devices -> Click the checkboxes for the 1080Ti (Graphics and Audio) and click on Bind to VFIO on Boot, this should make the card invisible for Unraid and the plugin so that ot ia reserved for exclusive access in the VM.

 

11 minutes ago, rragu said:

Install this nVidia plugin to use the P2000 for hardware transcoding in Docker

Exactly, follow the second post in this thread after you installed the plugin (you can assingn the P2000 in multiple Containers too).

5 minutes ago, ich777 said:

Go to Tools -> System Devices -> Click the checkboxes for the 1080Ti (Graphics and Audio) and click on Bind to VFIO on Boot, this should make the card invisible for Unraid and the plugin so that ot ia reserved for exclusive access in the VM.

 

Thanks!

 

Are there any possible issues that could occur as a result of stubbing the primary GPU (just wondering if there is something to look out for)?

1 hour ago, rragu said:

Are there any possible issues that could occur as a result of stubbing the primary GPU (just wondering if there is something to look out for)?

Not as far as I know but that would be something for the VM subforums.

But keep in mind that you don't have to extract the BIOS from the card anymore.

 

Try it first and if you experience any problems feel free to make a post again.

8 hours ago, ich777 said:

Please open up a console and can you give me the output from 'nvidia-smi'?

 

Can you also send me a secreenshot from the container template itself (with Advanced View turned on)?

 

Everything seems to look correct from configuration, unless I am missing something. 

Extra Parameters = --runtime=nvidia

Variable NVIDIA_VISIBLE_DEVICES = GPU UUID

Variable NVIDIA_DRIVER_CAPABILITIES = all

 

Also note that Tdar is currently transcoding with the GPU selected and there is no GPU activity using nvidia-smi.

23 minutes ago, exdox77 said:

Also note that Tdar is currently transcoding with the GPU selected and there is no GPU activity using nvidia-smi.

What kind of files are you transcoding? Are you sure that the 960 is capable of transcoding the file.

 

Keep in mind that the "older" Nvidia cards have problrms with some HEVC formats...

 

If it's a HEVC file then try a h264 file instead if that also doesn't work.

8 hours ago, ich777 said:

But keep in mind that you don't have to extract the BIOS from the card anymore.

Sorry, just to make sure I'm understanding you right:

I don't need to do anything to the 1080 Ti primary GPU other than bind it to VFIO via System Devices? I don't need to specify the vBIOS in the W10 VM's config/XML etc.?

30 minutes ago, rragu said:

I don't need to do anything to the 1080 Ti primary GPU other than bind it to VFIO via System Devices? I don't need to specify the vBIOS in the W10 VM's config/XML etc.?

No, but never say never...

Nvidia confirmed that they will add official support for Virtualization, please read this:

 

6 hours ago, ich777 said:

What kind of files are you transcoding? Are you sure that the 960 is capable of transcoding the file.

 

Keep in mind that the "older" Nvidia cards have problrms with some HEVC formats...

 

If it's a HEVC file then try a h264 file instead if that also doesn't work.

I figured it out, it was a config in one of the plugins that I needed to tell it to use the GPU. Working pretty well now.

I had issues a few pages ago, now I have again.
I rebooted the server, and when I started everything again, Plex wouldn't start and gave me then "bad parameter" error.
I checked it was typed in correctly, and it was.
I went to the "Nvidia Driver" page, and saw it didn't pick up the GPU. I tried verifying the checksum, and it came up with the error and downloaded it again. I did a hard refresh and checked it again, just to be sure.  It didn't pick up the GPU, so I rebooted it, again, just to be sure.
And it still doens't pick it up, and if I verify the checksum, it needs to download it again.

What am I doing wrong?

Edited June 7, 20215 yr by Nanobug

4 hours ago, Nanobug said:

I had issues a few pages ago, now I have again.
I rebooted the server, and when I started everything again, Plex wouldn't start and gave me then "bad parameter" error.

Yes, I remember that.

Can you try to issue the command 'nvidia-smi' and see what the output is? Or do you get again nothing?

 

4 hours ago, Nanobug said:

And it still doens't pick it up, and if I verify the checksum, it needs to download it again.

Please remember there was a driver update again a few days ago and eventually that file wasn't downloaded properly...

 

If this happens the next time please post your Diagnostics here and the output from nvidia-smi.

32 minutes ago, ich777 said:

Can you try to issue the command 'nvidia-smi' and see what the output is? Or do you get again nothing?

Nothing.
"bash: nvidia-smi: command not found"

 

32 minutes ago, ich777 said:

Please remember there was a driver update again a few days ago and eventually that file wasn't downloaded properly...

 

If this happens the next time please post your Diagnostics here and the output from nvidia-smi.

It gave me a notification about the driver, I updated it, it didn't show anything, So I rebooted again, and it gave me the same answer again.

nvidia-smi:
"bash: nvidia-smi: command not found"

 

Diagnostics:
nanostorage-diagnostics-20210607-1342.zip
 

Edited June 7, 20215 yr by Nanobug

1 hour ago, Nanobug said:

Nothing.
"bash: nvidia-smi: command not found"

Then the driver download failed or is not finished successfully but that is nearly impossible because it checks the downloaded file against the MD5 sum.

 

Can you give me the output from:

ls -la /boot/config/plugins/nvidia-driver/packages/

and also from:

md5sum /boot/config/plugins/nvidia-driver/packages/*

 

34 minutes ago, ich777 said:

Can you give me the output from:

ls -la /boot/config/plugins/nvidia-driver/packages/

 

34 minutes ago, ich777 said:

and also from:

md5sum /boot/config/plugins/nvidia-driver/packages/*

2 hours ago, Nanobug said:

 

The md5 sum matches, but I think something does went really wrong on your system when the plugin installs the driver can you try to do a:

installpkg /boot/config/plugins/nvidia-driver/packages/nvidia-460.84.5.10.28-Unraid-1.txz

 

If that succeeds please try to issue the command nvidia-smi once more, if that also succeeds then try to stop and start the Docker service and see if Plex is working again after you stopped it.

 

This is a complete stock Unraid 6.9.2 build you've said or am I wrong?

26 minutes ago, ich777 said:

The md5 sum matches, but I think something does went really wrong on your system when the plugin installs the driver can you try to do a:

installpkg /boot/config/plugins/nvidia-driver/packages/nvidia-460.84.5.10.28-Unraid-1.txz

 

If that succeeds please try to issue the command nvidia-smi once more, if that also succeeds then try to stop and start the Docker service and see if Plex is working again after you stopped it.

 

This is a complete stock Unraid 6.9.2 build you've said or am I wrong?

This is odd....

- I think it's a typo,

 

This worked:

 

nvidia-smi gave me this now:

- the driver page gave it as well:

- There's the unRAID version. I haven't done anything but updating when it came out.
 

 

5 minutes ago, Nanobug said:

- I think it's a typo,

I think that's the case.

 

Run:

depmod -a

modprobe nvidia

 

After that go to the plugin page again.

8 minutes ago, ich777 said:

Run:

depmod -a

modprobe nvidia

 

After that go to the plugin page again.

It's there now:

 

Plex is not happy:

I tried with the GPUID as well, same result.

I've already contacted LSIO about it, but if you have the solution, I'd be happy too

3 minutes ago, Nanobug said:

I've already contacted LSIO about it, but if you have the solution, I'd be happy too

Have you restarted the Docker service too?

30 minutes ago, ich777 said:

Have you restarted the Docker service too?

I can't even start it

8 minutes ago, Nanobug said:

I can't even start it

I don't understand, what do you mean with you can't start it?

 

Disable the whole Docker service and then enable it again: