Alder lake GPU support in scale?

[tom9933]

Hello,
I recently built a new server with an Intel i7-12700K Alder Lake processor. Before building this server I tested scale (RC1) on an older system (6th gen intel) and was excited to see that hardware transcoding in plex just worked (core took some tweaking..). Now I have the latest version of scale running on the new server, and everything is working great with the exception of the GPU setting in apps. I have tried both the official plex app and the truecharts version. While I have GPU checkbox enabled under the advanced app section I don't see the dropdown to choose it under the apps.

I found another post that talked about running a command (midclt call device.get_gpus | jq .) and the output is below. I'm also running two windows VMs, but I do NOT have GPU pass thru enabled on those VMS. I did select it from the drop down on the VM device configuration and then I see the following which seems odd.

Is my system just too new or could it be a configuration issue?
Thanks Tom

root@truenas[~]# midclt call device.get_gpus | jq .
[
{
"addr": {
"pci_slot": "0000:00:02.0",
"domain": "0000",
"bus": "00",
"slot": "02"
},
"description": "Intel Corporation AlderLake-S GT1",
"devices": [
{
"pci_id": "8086:4668",
"pci_slot": "0000:00:00.0",
"vm_pci_slot": "pci_0000_00_00_0"
},
{
"pci_id": "8086:4680",
"pci_slot": "0000:00:02.0",
"vm_pci_slot": "pci_0000_00_02_0"
},
{
"pci_id": "8086:467D",
"pci_slot": "0000:00:0a.0",
"vm_pci_slot": "pci_0000_00_0a_0"
},
{
"pci_id": "8086:467F",
"pci_slot": "0000:00:0e.0",
"vm_pci_slot": "pci_0000_00_0e_0"
},
{
"pci_id": "8086:7AE0",
"pci_slot": "0000:00:14.0",
"vm_pci_slot": "pci_0000_00_14_0"
},
{
"pci_id": "8086:7AA7",
"pci_slot": "0000:00:14.2",
"vm_pci_slot": "pci_0000_00_14_2"
},
{
"pci_id": "8086:7AF0",
"pci_slot": "0000:00:14.3",
"vm_pci_slot": "pci_0000_00_14_3"
},
{
"pci_id": "8086:7ACC",
"pci_slot": "0000:00:15.0",
"vm_pci_slot": "pci_0000_00_15_0"
},
{
"pci_id": "8086:7ACD",
"pci_slot": "0000:00:15.1",
"vm_pci_slot": "pci_0000_00_15_1"
},
{
"pci_id": "8086:7ACE",
"pci_slot": "0000:00:15.2",
"vm_pci_slot": "pci_0000_00_15_2"
},
{
"pci_id": "8086:7AE8",
"pci_slot": "0000:00:16.0",
"vm_pci_slot": "pci_0000_00_16_0"
},
{
"pci_id": "8086:7AE2",
"pci_slot": "0000:00:17.0",
"vm_pci_slot": "pci_0000_00_17_0"
},
{
"pci_id": "8086:7AC8",
"pci_slot": "0000:00:1a.0",
"vm_pci_slot": "pci_0000_00_1a_0"
},
{
"pci_id": "1CC1:8201",
"pci_slot": "0000:01:00.0",
"vm_pci_slot": "pci_0000_01_00_0"
},
{
"pci_id": "8086:7AC0",
"pci_slot": "0000:00:1b.0",
"vm_pci_slot": "pci_0000_00_1b_0"
},
{
"pci_id": "8086:7AC4",
"pci_slot": "0000:00:1b.4",
"vm_pci_slot": "pci_0000_00_1b_4"
},
{
"pci_id": "1000:0072",
"pci_slot": "0000:03:00.0",
"vm_pci_slot": "pci_0000_03_00_0"
},
{
"pci_id": "8086:7AB8",
"pci_slot": "0000:00:1c.0",
"vm_pci_slot": "pci_0000_00_1c_0"
},
{
"pci_id": "8086:7ABA",
"pci_slot": "0000:00:1c.2",
"vm_pci_slot": "pci_0000_00_1c_2"
},
{
"pci_id": "10EC:8125",
"pci_slot": "0000:05:00.0",
"vm_pci_slot": "pci_0000_05_00_0"
},
{
"pci_id": "8086:7AB0",
"pci_slot": "0000:00:1d.0",
"vm_pci_slot": "pci_0000_00_1d_0"
},
{
"pci_id": "8086:7AB4",
"pci_slot": "0000:00:1d.4",
"vm_pci_slot": "pci_0000_00_1d_4"
},
{
"pci_id": "1CC1:8201",
"pci_slot": "0000:07:00.0",
"vm_pci_slot": "pci_0000_07_00_0"
},
{
"pci_id": "8086:7A84",
"pci_slot": "0000:00:1f.0",
"vm_pci_slot": "pci_0000_00_1f_0"
},
{
"pci_id": "8086:7AD0",
"pci_slot": "0000:00:1f.3",
"vm_pci_slot": "pci_0000_00_1f_3"
},
{
"pci_id": "8086:7AA3",
"pci_slot": "0000:00:1f.4",
"vm_pci_slot": "pci_0000_00_1f_4"
},
{
"pci_id": "8086:7AA4",
"pci_slot": "0000:00:1f.5",
"vm_pci_slot": "pci_0000_00_1f_5"
}
],
"vendor": "INTEL",
"available_to_host": true
}
]

 

[ClassicGOD]

The drop down while creating an app is a bug I think. Try installing the app without selecting the GPU then editing it. I had a few times where my Quadro P1000 was unavailable when installing the app but I was able to edit the app and select it there. After that HW transcoding worked as intended.

As for the other message - to pass GPU to a VM you need to isolate it. Isolating means preventing the host OS from loading any driver and initializing the device. Like the message says - OS needs at least 1 grapics adapter for itself and you have no GPUs isolated at this time (you can set them up in System Settings > Advanced. If i remember correctly TrueNas will attempt to set it up for you "automagically" when you try to pass GPU to a VM.).

You don't need to isolate a GPU for usage inside of a container/app which makes the entire HW acceleration in PLEX much easier and less complex.

 

[tom9933]

So while I see the GPU checkbox in the advanced (all apps) setting, I don't see anything in any new apps. It lists GPU but I don't see the dropdown or any other settings. When I toggle the advanced checkbox I do see the GPU section removed in the instance but it never gives me a configuration option. I confirmed that plex is NOT using hardware transcoding with a quick playback test.

Its odd though because in the GPU isolation page (not enabled) I do see it with a drop down as expected.

 

[ClassicGOD]

So while I see the GPU checkbox in the advanced (all apps) setting, I don't see anything in any new apps. It lists GPU but I don't see the dropdown or any other settings. When I toggle the advanced checkbox I do see the GPU section removed in the instance but it never gives me a configuration option. I confirmed that plex is NOT using hardware transcoding with a quick playback test.

Try editing already running Plex app. Do you have an option for GPU there?

 

[tom9933]

Unfortunately no with a new app or editing an existing one all I see is this. Now this does go away/return when toggling the checkbox under advanced options so it seems like the system just doesn't see my GPU (quicksync) being available for an app.

I wonder if one of these docker images needs an upgrade?

 

[tom9933]

So I did a little looking and found that indeed that plugin version is older. I was able to pull the updated docker images using commands through the shell. Nothing changed so I then did a reboot and still nothing changed. I'm tempted to delete the .19 versions but I dont want to break something.. I also wonder about the unnamed docker images...

 

[ClassicGOD]

Those are images for otter support containers running in the background not visible in ui. You can view them from cli with "docker ps" and get details with "docker inspect <id>"

At this point nothing else comes to mind other than create a JIRA ticket on https://jira.ixsystems.com/ and wait for answers. Or test RC2 that's is going to be released hopefully tomorrow.

 

[truecharts]

Important note:
Full AlderLike support has landed between kernel 5.14 and... well... it's not completely stable yet.

SCALE runs 5.10
So it's best to simply don't expect alderlake to be compatible and be happy when something works.

 

[tom9933]

Important note:
Full AlderLike support has landed between kernel 5.14 and... well... it's not completely stable yet.

SCALE runs 5.10
So it's best to simply don't expect alderlake to be compatible and be happy when something works.

Thank you for the confirmation!!! I thought that might be the case but I was having a hard time finding it documented anywhere.