Question on intel arc a380 support wrt plex

[Aaron Gould]

Reportedly, passing the iGPU to Plex requires a C2x6 chipset, not C2x2.

This is what I've seen in a forum post on Reddit. I can't seem to get a solid answer on whether a discrete GPU (e.g. Intel Arc) would get around this limitation though.

 

[HoneyBadger]

Reportedly, passing the iGPU to Plex requires a C2x6 chipset, not C2x2.

Unless I'm misunderstanding, @Aaron Gould is talking about adding an ARC A380 add-in card and passing that to a VM as a PCIe device. The C236 requirement is for simultaneous BMC+iGPU enablement, and using the iGPU for Apps (which is still considered a host-GPU workload)

The success/failure would probably be determined on the board's ability to keep the onboard BMC's ASPEED AST2400 enabled and set as the "primary video device" after the A380 is installed. The manual seems non-deterministic about that though:

Graphics Configuration
Primary Display
Use this feature to select a IGFX/PE/PCI graphics device to be used as the primary display.
You can also select SG to use a Switchable Gfx device for your primary display.
The options are Auto, PEG, and PCIE.

Primary PCIE (PCI-Express Graphics)
This feature allows the user to specify which graphics card to be used as the primary
graphics device. The options are Onboard, PCH SLOT4 PCI-E 3.0 x1, and PCH SLOT5
PCI-E 3.0 x4(IN X8).

 

[Etorix]

Unless I'm misunderstanding, @Aaron Gould is talking about adding an ARC A380 add-in card and passing that to a VM as a PCIe device. The C236 requirement is for simultaneous BMC+iGPU enablement, and using the iGPU for Apps (which is still considered a host-GPU workload)

Correct, although the question comes in part from @Aaron Gould resurrecting an older thread . But as he reported on the failure of his C232 platform to passthrough the (i)GPU, I wasn't sure if your answer about "GPU" was about a possible Arc dGPU, which should work, or about the iGPU, which doesn't (although it should still be another VT-d thing, but, hey, Intel wants you to buy a C236).

 

[Aaron Gould]

Thanks guys. I decided to just take a leap of faith and ordered an Arc to be delivered tomorrow. Here's hoping it works and isn't too much trouble to get working!

 

[HoneyBadger]

Thanks guys. I decided to just take a leap of faith and ordered an Arc to be delivered tomorrow. Here's hoping it works and isn't too much trouble to get working!

Keep us posted on the results! The AV1 speed on even the entry level ARC cards is quite good.

 

[Aaron Gould]

Got the Arc delivered today... and immediately encountered a problem. The card is too long and the SAS3 plugs are blocking installation.

I can't use the 6 SATA3 on-board plugs because I have 6 drives plus a SATA DOM.

My own fault for not checking dimensions. Guess I have some re-thinking to do.

 

[skittlebrau]

Unless I'm misunderstanding, @Aaron Gould is talking about adding an ARC A380 add-in card and passing that to a VM as a PCIe device. The C236 requirement is for simultaneous BMC+iGPU enablement, and using the iGPU for Apps (which is still considered a host-GPU workload)

The success/failure would probably be determined on the board's ability to keep the onboard BMC's ASPEED AST2400 enabled and set as the "primary video device" after the A380 is installed. The manual seems non-deterministic about that though:

Another caveat is also IOMMU groups. Some boards have crappy IOMMU group separation and won't allow even CPU-connected PCI-E slots from being isolated, as is the case for my Gigabyte C246M-WU4 Xeon board, and that prevents GPU passthrough as well.

 

[Aaron Gould]

Another caveat is also IOMMU groups. Some boards have crappy IOMMU group separation and won't allow even CPU-connected PCI-E slots from being isolated, as is the case for my Gigabyte C246M-WU4 Xeon board, and that prevents GPU passthrough as well.

A bit later than I was hoping to get to this. I ended up returning the A750 and got an A380 instead. Much lower power draw for my purposes (225W vs 75W).

Now I'm wondering if my SuperMicro X11SSL-CF might be one of these boards affected by crappy IOMMU group separation. Here's the error that appears when firing up a VM with the GPU passed through:

internal error: qemu unexpectedly closed the monitor: 2024-01-09T03:52:53.161694Z qemu-system-x86_64: -device {"driver":"vfio-pci","host":"0000:03:00.0","id":"hostdev0","bus":"pci.0","addr":"0x6"}: vfio 0000:03:00.0: group 1 is not viable Please ensure all devices within the iommu_group are bound to their vfio bus driver.

Here's what's in "group 1", which my VM is complaining

00:01.0 PCI bridge [0604]: Intel Corporation 6th-10th Gen Core Processor PCIe Controller (x16) [8086:1901] (rev 07) 00:01.1 PCI bridge [0604]: Intel Corporation Xeon E3-1200 v5/E3-1500 v5/6th Gen Core Processor PCIe Controller (x8) [8086:1905] (rev 07) 01:00.0 PCI bridge [0604]: Intel Corporation Device [8086:4fa1] (rev 01) 02:01.0 PCI bridge [0604]: Intel Corporation Device [8086:4fa4] 02:04.0 PCI bridge [0604]: Intel Corporation Device [8086:4fa4] 03:00.0 VGA compatible controller [0300]: Intel Corporation DG2 [Arc A380] [8086:56a5] (rev 05) 04:00.0 Audio device [0403]: Intel Corporation DG2 Audio Controller [8086:4f92] 05:00.0 Serial Attached SCSI controller [0107]: Broadcom / LSI SAS3008 PCI-Express Fusion-MPT SAS-3 [1000:0097] (rev 02)

Am I screwed until (and if) Cobia's kernel is upgraded to 6.2?

 

[skittlebrau]

Cobia will be staying in kernel 6.1 I think, unless I’ve misinterpreted the release schedule.

You won’t need to wait long for Dragonfish 24.04 to be released however. That may bump up the kernel version.

In the interim, I’ve been able to force the i915 driver to load via modprobe in Grub so my Arc A380 works in apps.

 

[Aaron Gould]

In the interim, I’ve been able to force the i915 driver to load via modprobe in Grub so my Arc A380 works in apps.

I've done some Googling, and found a few hits on adding this to my Grub. In /etc/default/grub.d, I appended " nomodeset i915.modeset=0" to the GRUB_CMDLINE_LINUX_DEFAULT line.

I also issued the following command (as root):

midclt call system.advanced.update '{"kernel_extra_options": "nomodeset i915.modeset=0"}'

After a reboot, I'm still not seeing the GPU available to my Jellyfin app.

Seems to be present in list of VGA devices:

root@truenas[~]# lspci|grep -i vga 00:13.0 Non-VGA unclassified device: Intel Corporation 100 Series/C230 Series Chipset Family Integrated Sensor Hub (rev 31) 03:00.0 VGA compatible controller: Intel Corporation DG2 [Arc A380] (rev 05) 09:00.0 VGA compatible controller: ASPEED Technology, Inc. ASPEED Graphics Family (rev 30)

Were there other/alternative steps you had to do?

 

[skittlebrau]

Sorry, I was on my phone at the time and didn't have the command handy.

While this post is in reference to forcing Intel iGPU drivers for newer CPUs to load, the steps are the same for forcing the experimental discrete Arc GPU driver in the 6.1 kernel to load since they also use the i915 driver.

PSA: How to Fix Quick Sync on newer Intel CPUs in SCALE

Okay, so this took awhile to figure out, and I want to thank @waqarahmed for giving me the command that ultimately got my Quick Sync working in TrueNAS SCALE and let me pass it through to my Apps. This post does not apply for those with nvidia cards. It only applies to those of us who...

www.truenas.com

Don't just blindly use the midclt command in that thread however. You must find the correct PCI vendor ID relevant for your GPU, which you can find by using:

Code:

lspci -nn

In my particular case with my AsRock A380 Challenger ITX GPU, I used the command:

Code:

sudo midclt call system.advanced.update '{"kernel_extra_options": "i915.force_probe=56a5"}'

After doing that, my Arc GPU was available to use in apps.

 

[Aaron Gould]

Don't just blindly use the midclt command in that thread however. You must find the correct PCI vendor ID relevant for your GPU, which you can find by using:

Code:

lspci -nn

In my particular case with my AsRock A380 Challenger ITX GPU, I used the command:

Code:

sudo midclt call system.advanced.update '{"kernel_extra_options": "i915.force_probe=56a5"}'

After doing that, my Arc GPU was available to use in apps.

Thank you, this has appeared to solve the hardware issue. Mine is also "56a5" (probably not surprising given we both have the same A380 card). I updated the configs, rebooted, and my app can now be assigned a GPU in the resources section.

I'm getting a failed playback with 60-70% of the movies I've tried, but I can dig into that now that the hardware side is behind me!

 

[Gizmmo]

Hi guys, thanks for the guide. I managed to install my arc380 and managed to pass 5 cores to Jellyfin buy transcoding doesn't work.
I noticed the GPU dropdown is gone and was replaced by an input field called "Add Intel i915 GPUs"? I noticed this even before i installed my arc380. When i was using my iGPU, i typed 3 in the input and transcoding worked. Still didn't have a GPU dropdown like in the past.
Is this normal or i should see a dropdown?