SimonF

You can do at the PCI device level also in system devices by clicking on the warning triangle. I added the feature to stop VMs starting if new hardware has been added an the VMs have not been updated as this can cause issues of correcting the wrong device to a VM, i.e. the HBA which would mean the disks would disappear.

Yes should be blacklist by default.

Docker folders is likely to be the cause.

Could you create a new thread and post diagnostics so it can be review to see If additional parts are needed.

Intel GPU top does not work with XE driver needed for the card. Have a look at this thread. https://forums.unraid.net/topic/193523-hardware-transcode-i915-driver-vs-xe/#findComment-1578648 You may need to enable XE driver for you card.

root@computenode:~# cat /usr/src/linux-6.12.47-Unraid/.config | grep IVPU CONFIG_DRM_ACCEL_IVPU=m The module Is there but maybe PCI ID is not in the 6.12 kernel and will need a later version.

My NPU show no drivers modules in 7.2beta lspci -ks 0000:80:14.5 80:14.5 Non-VGA unclassified device: Intel Corporation Device 7f2f (rev 10) Subsystem: Micro-Star International Co., Ltd. [MSI] Device 7e17

Install this dev release, https://raw.githubusercontent.com/SimonFair/gpustat-unraid/refs/heads/Dynamic-App-Support/gpustat.plg it should have the fix above from @bgo82

Yes that is what I use in gpustat. I am working on other options for XE, some are kernel dependant. Utilisation is not provided as far as I can see at present.

Intel_gpu_top only works with the i915 driver so will not find the cards using XE. cat /sys/kernel/debug/dri/$pciId/clients should show the clients using the gpu replace $pciid witht the card IP. gpustat has some basic support for XE to show apps but no util etc.

Unraid is on Kernel 6.12 looks like GPU needs at least 6.14, sr-iov support is not available as yet. Intel's Linux driver guidance for the Arc Pro B50 at launch is to be using Ubuntu 25.04 with the Linux 6.14 kernel and Mesa 25.0.7 or newer. Linux 6.14+ and Mesa 25.0+ is easy to meet by most modern Linux distributions. Of course, the newer the kernel and Mesa drivers, the better the experience in terms of features and performance. As noted there is a lot in the current Linux 6.17 development tree around Project Battlematrix preparations and more for SR-IOV, multi-device prep, etc. The newer Mesa 25.1/25.2 and 25.3-devel releases also have additional performance optimizations and capabilities for Intel Battlemage GPUs at large for both the Iris Gallium3D and ANV Vulkan drivers.

Maybe not sure, can you try without the splitter as a test?

If you run this command do you get any errors? root@QNAPR420:~# corsairmi name: 'CORSAIR HX1000i PSU' vendor: 'CORSAIR HX1000i PSU' product: 'CORSAIR HX1000i PSU' powered: 2392322 (27d. 16h) uptime: 51722 (0d. 14h) temp1: 47.8 temp2: 56.5 fan rpm: 0.0 supply volts: 230.0 total watts: 132.0 output0 volts: 12.1 output0 amps: 8.0 output0 watts: 96.0 output1 volts: 5.0 output1 amps: 5.7 output1 watts: 28.5 output2 volts: 3.3 output2 amps: 2.9 output2 watts: 9.5

You need to stop the i915 driver binding to the device and get xe to load for it. From lspcu you can find the info required for the force_probe. lspci -nn | grep VGA 00:02.0 VGA compatible controller [0300]: Intel Corporation Arrow Lake-S [Intel Graphics] [8086:7d67] (rev 06) 04:00.0 VGA compatible controller [0300]: Intel Corporation DG2 [Arc A770] [8086:56a0] (rev 08) 0e:00.0 VGA compatible controller [0300]: NVIDIA Corporation TU106 [GeForce RTX 2060 Rev. A] [10de:1f08] (rev a1) 80:14.5 Non-VGA unclassified device [0000]: Intel Corporation Device [8086:7f2f] (rev 10) in my case 56a0 for the A770 For the i915 driver you need to do options i915 force_probe=!56a0 note ! in front of the model. and for XE options xe force_probe=56a0 You can add them to modprobe files via system drivers in tools as below, but change to match your device.

Have a look here to see if that is supported. http://lcdproc.omnipotent.net/hardware.php3

Which versions of OS and Plugin are you running. Did you create any diagnostics? Maybe increase refresh timer

Should be fine.

Is the image ovmf or do you need seabios?

It is because the way the card reports itself in sysfs, it is a cosmestic issue, the card will be running at faster than indicated, when looking at the nvtop code they look upstream to get a more realistic value I will see if I can do that.

No this is for older gen CPUs. B50 will be sriov but is not currently in kernel used in Unraid, I think it is in 6.17 which is still in RC.

Will look a source code from nvtop to see how they calculate.

Maybe post unraid diagnostics. If you disks are spun down then spin them up and nut diags should work.

You can stay on the Dev release. I will update to use the prod stream when issued.

Yes if I comment out the line runtime does not show in the output.

Thanks for the help. This server rarely gets rebooted so will leave patched driver for now.