Equipment:Z790+5060ti

OS: Unraid7.3.3-rc2

The configuration details of virtual machine Win11 can be found in the XML file. In Unriad7.3.2, everything is working normally. When upgrading to Unraid 7.3.3-rc2 or Unraid 7.4-beta3, the virtual machine starts with an error. How can I solve the startup error? Background: Z790+5060Ti, for virtual machine Win11, simply connect to the Nvidia sound card. After turning off the virtual machine Win11, Unraid's webui freezes and spins, causing a malfunction in the virtual machine's functionality. It can only be shut down by forcibly pressing the switch. Therefore, physical sound cards cannot be used, only virtual sound cards can be used

lynxserver-diagnostics-20260928-1658.zip win11 xml.txt syslog vm.txt

Thanks for the diagnostics. The VM log shows the cause:

vfio 0000:01:00.0: group 18 is not viable. Please ensure all devices within the iommu_group are bound to their vfio bus driver.

Your 5060 Ti's graphics device (01:00.0) and its HDMI audio device (01:00.1) are in the same IOMMU group. The VM passes through only 01:00.0. The audio device stays with the host sound driver, so VFIO will not start the VM. This is a configuration problem, not a 7.3.3 regression. Most likely your 7.3.2 setup still passed the NVIDIA audio through.

You can keep the virtual sound card. Do one of these:

1. Go to Tools > System Devices. Check only the NVIDIA audio device (01:00.1, 10de:22eb), click Bind selected to VFIO at boot, and reboot. The GPU can still move between the host and the VM.

2. If you don't need the GPU in Unraid or Docker, bind both 01:00.0 and 01:00.1 to VFIO at boot and reboot. This also avoids a known NVIDIA driver problem when the GPU returns to the host after the VM stops. That problem is a likely cause of the WebGUI freeze you had after shutting down the VM with the NVIDIA sound card passed through.

If you try passing the NVIDIA audio through again and the WebGUI still freezes on VM shutdown, please post the diagnostics. Because the server needs a hard reset, first enable the syslog server (mirror to flash) so we can see the log from the freeze.

19 minutes ago, JorgeB said:

谢谢你的诊断。VM日志显示了原因：

vfio 0000:01:00.0: group 18 is not viable. Please ensure all devices within the iommu_group are bound to their vfio bus driver.

你的5060 Ti的显卡设备（01：00.0）和HDMI音频设备（01：00.1）属于同一个IOMMU组。虚拟机只通过01：00.0。音频设备会保留在主机的声音驱动上，所以VFIO不会启动虚拟机。这是配置问题，不是7.3.3的回归问题。很可能你的7.3.2系统还能通过NVIDIA音频传输。

你可以保留虚拟声卡。做以下其中之一：

1. 进入工具>系统设备。只检查NVIDIA音频设备（01：00.1,10de：22eb），在启动时点击“绑定”选择VFIO，然后重启。GPU仍然可以在主机和虚拟机之间移动。

2. 如果你不需要在Unraid或Docker里使用显卡，开机时将01：00.0和01：00.1绑定到VFIO，然后重启。这也避免了虚拟机停止后GPU返回主机时已知的NVIDIA驱动问题。这个问题很可能是你关闭虚拟机并插入NVIDIA声卡后WebGUI卡死的原因。

如果你再次尝试传递NVIDIA音频，但虚拟机关机时WebGUI依然卡死，请发布诊断结果。因为服务器需要硬重置，首先启用syslog服务器（镜像到闪存），这样我们才能看到冻结时的日志。

I need to use an Nvidia graphics card in Unraid and Docker, and the virtual machine can be directly connected to the Nvidia card. The virtual machine can start normally. What should I do next? Thank you

1 hour ago, JorgeB said:

Thanks for the diagnostics. The VM log shows the cause:

vfio 0000:01:00.0: group 18 is not viable. Please ensure all devices within the iommu_group are bound to their vfio bus driver.

Your 5060 Ti's graphics device (01:00.0) and its HDMI audio device (01:00.1) are in the same IOMMU group. The VM passes through only 01:00.0. The audio device stays with the host sound driver, so VFIO will not start the VM. This is a configuration problem, not a 7.3.3 regression. Most likely your 7.3.2 setup still passed the NVIDIA audio through.

You can keep the virtual sound card. Do one of these:

1. Go to Tools > System Devices. Check only the NVIDIA audio device (01:00.1, 10de:22eb), click Bind selected to VFIO at boot, and reboot. The GPU can still move between the host and the VM.

2. If you don't need the GPU in Unraid or Docker, bind both 01:00.0 and 01:00.1 to VFIO at boot and reboot. This also avoids a known NVIDIA driver problem when the GPU returns to the host after the VM stops. That problem is a likely cause of the WebGUI freeze you had after shutting down the VM with the NVIDIA sound card passed through.

If you try passing the NVIDIA audio through again and the WebGUI still freezes on VM shutdown, please post the diagnostics. Because the server needs a hard reset, first enable the syslog server (mirror to flash) so we can see the log from the freeze.

1.When I try to select only the NVIDIA audio device (01:00.1, 10de:22eb), the 01:00.0 VGA (RTX 5060 Ti) gets checked automatically. I cannot check the NVIDIA audio device alone Bind selected to VFIO at boot

1 hour ago, JorgeB said:

Thanks for the diagnostics. The VM log shows the cause:

vfio 0000:01:00.0: group 18 is not viable. Please ensure all devices within the iommu_group are bound to their vfio bus driver.

Your 5060 Ti's graphics device (01:00.0) and its HDMI audio device (01:00.1) are in the same IOMMU group. The VM passes through only 01:00.0. The audio device stays with the host sound driver, so VFIO will not start the VM. This is a configuration problem, not a 7.3.3 regression. Most likely your 7.3.2 setup still passed the NVIDIA audio through.

You can keep the virtual sound card. Do one of these:

1. Go to Tools > System Devices. Check only the NVIDIA audio device (01:00.1, 10de:22eb), click Bind selected to VFIO at boot, and reboot. The GPU can still move between the host and the VM.

2. If you don't need the GPU in Unraid or Docker, bind both 01:00.0 and 01:00.1 to VFIO at boot and reboot. This also avoids a known NVIDIA driver problem when the GPU returns to the host after the VM stops. That problem is a likely cause of the WebGUI freeze you had after shutting down the VM with the NVIDIA sound card passed through.

If you try passing the NVIDIA audio through again and the WebGUI still freezes on VM shutdown, please post the diagnostics. Because the server needs a hard reset, first enable the syslog server (mirror to flash) so we can see the log from the freeze.

7.3.2

1 hour ago, JorgeB said:

Thanks for the diagnostics. The VM log shows the cause:

vfio 0000:01:00.0: group 18 is not viable. Please ensure all devices within the iommu_group are bound to their vfio bus driver.

Your 5060 Ti's graphics device (01:00.0) and its HDMI audio device (01:00.1) are in the same IOMMU group. The VM passes through only 01:00.0. The audio device stays with the host sound driver, so VFIO will not start the VM. This is a configuration problem, not a 7.3.3 regression. Most likely your 7.3.2 setup still passed the NVIDIA audio through.

You can keep the virtual sound card. Do one of these:

1. Go to Tools > System Devices. Check only the NVIDIA audio device (01:00.1, 10de:22eb), click Bind selected to VFIO at boot, and reboot. The GPU can still move between the host and the VM.

2. If you don't need the GPU in Unraid or Docker, bind both 01:00.0 and 01:00.1 to VFIO at boot and reboot. This also avoids a known NVIDIA driver problem when the GPU returns to the host after the VM stops. That problem is a likely cause of the WebGUI freeze you had after shutting down the VM with the NVIDIA sound card passed through.

If you try passing the NVIDIA audio through again and the WebGUI still freezes on VM shutdown, please post the diagnostics. Because the server needs a hard reset, first enable the syslog server (mirror to flash) so we can see the log from the freeze.

The VM was configured this way on 7.3.2: only the NVIDIA GPU was passed through, without passing through the HDMI audio device. If I roll back to 7.3.2 now, the VM boots normally. I have confirmed that 7.3.2 works fine with no configuration changes. After upgrading to either 7.3.3rc2 or 7.4beta3, the VM fails to start

Thanks for clarifying. I was too quick to rule out a regression. Your rollback test, with no configuration changes, is important evidence.

You are also correct about System Devices selecting both functions together. My instruction to select only the NVIDIA audio device there was incorrect.

Since you need the GPU for Unraid and Docker, binding both functions to VFIO at boot would not meet your requirements. Keep your virtual audio configuration for now.

Please upload diagnostics from 7.3.2 while this VM is running successfully, together with its VM XML. We already have the failed-start diagnostics from 7.3.3-rc.2.

This will let us compare the IOMMU groups, the audio device’s driver, and the software versions between the working and failing setups. The current error explains why VFIO rejects startup, but we still need to establish why the same configuration works on 7.3.2.

The earlier shutdown freeze also remains unexplained. There is no need to repeat the physical-audio passthrough test at this stage.

19 minutes ago, JorgeB said:

Thanks for clarifying. I was too quick to rule out a regression. Your rollback test, with no configuration changes, is important evidence.

You are also correct about System Devices selecting both functions together. My instruction to select only the NVIDIA audio device there was incorrect.

Since you need the GPU for Unraid and Docker, binding both functions to VFIO at boot would not meet your requirements. Keep your virtual audio configuration for now.

Please upload diagnostics from 7.3.2 while this VM is running successfully, together with its VM XML. We already have the failed-start diagnostics from 7.3.3-rc.2.

This will let us compare the IOMMU groups, the audio device’s driver, and the software versions between the working and failing setups. The current error explains why VFIO rejects startup, but we still need to establish why the same configuration works on 7.3.2.

The earlier shutdown freeze also remains unexplained. There is no need to repeat the physical-audio passthrough test at this stage.

Thank you. Here are the diagnostics from 7.3.2 and the VM XML. The Windows 11 VM is currently booted and working normally.

win11 xml.txt lynxserver-diagnostics-20260928-2101.zip

Thanks, these diagnostics show why the same VM configuration works on 7.3.2.

On 7.3.2, the NVIDIA audio device has no driver bound. On 7.3.3-rc.2, the host claims it with snd_hda_intel. Because it shares an IOMMU group with the GPU, that prevents VFIO from starting the VM. Your uploaded VM XMLs match.

Please try this temporary test on 7.3.3-rc.2. Leave the VM configuration and virtual audio unchanged.

With the VM stopped, stop any containers using the GPU. Make sure the host is not using the card’s HDMI/DisplayPort audio.

In the Unraid terminal, run:

lspci -nnk -s 01:00.1

If it shows Kernel driver in use: snd_hda_intel, run:

echo 0000:01:00.1 > /sys/bus/pci/drivers/snd_hda_intel/unbind

lspci -nnk -s 01:00.1

If the command reports an error, stop and post that error. Otherwise, confirm that the audio device no longer shows a Kernel driver in use line, then start the VM.

This temporarily releases only the NVIDIA audio device from the host driver. It makes no boot configuration changes, and a host reboot restores normal driver probing.

Please report whether the VM starts. If it still fails, attach fresh diagnostics. This test addresses startup only; the earlier shutdown freeze remains unexplained.

6 minutes ago, JorgeB said:

Thanks, these diagnostics show why the same VM configuration works on 7.3.2.

On 7.3.2, the NVIDIA audio device has no driver bound. On 7.3.3-rc.2, the host claims it with snd_hda_intel. Because it shares an IOMMU group with the GPU, that prevents VFIO from starting the VM. Your uploaded VM XMLs match.

Please try this temporary test on 7.3.3-rc.2. Leave the VM configuration and virtual audio unchanged.

With the VM stopped, stop any containers using the GPU. Make sure the host is not using the card’s HDMI/DisplayPort audio.

In the Unraid terminal, run:

lspci -nnk -s 01:00.1

If it shows Kernel driver in use: snd_hda_intel, run:

echo 0000:01:00.1 > /sys/bus/pci/drivers/snd_hda_intel/unbind

lspci -nnk -s 01:00.1

If the command reports an error, stop and post that error. Otherwise, confirm that the audio device no longer shows a Kernel driver in use line, then start the VM.

This temporarily releases only the NVIDIA audio device from the host driver. It makes no boot configuration changes, and a host reboot restores normal driver probing.

Please report whether the VM starts. If it still fails, attach fresh diagnostics. This test addresses startup only; the earlier shutdown freeze remains unexplained.

Okay, I will boot into 7.3.3rc2 and perform this test. Thank you.

16 minutes ago, HDLynx said:

Okay, I will boot into 7.3.3rc2 and perform this test. Thank you.

root@LynxServer:~# lspci -nnk -s 01:00.1

01:00.1 Audio device [0403]: NVIDIA Corporation GB206 High Definition Audio Controller [10de:22eb] (rev a1)

Subsystem: NVIDIA Corporation Device [10de:0000]

Kernel driver in use: snd_hda_intel

Kernel modules: snd_hda_intel

root@LynxServer:~# echo 0000:01:00.1 > /sys/bus/pci/drivers/snd_hda_intel/unbind

root@LynxServer:~# lspci -nnk -s 01:00.1

01:00.1 Audio device [0403]: NVIDIA Corporation GB206 High Definition Audio Controller [10de:22eb] (rev a1)

Subsystem: NVIDIA Corporation Device [10de:0000]

Kernel modules: snd_hda_intel

After executing the unbind command echo 0000:01:00.1 > /sys/bus/pci/drivers/snd_hda_intel/unbind, the Windows 11 VM on 7.3.3-rc2 starts and runs perfectly. This same problem occurs on 7.3.3-rc2 as well as 7.4-beta3. Can this be fixed? Or will I have to manually unbind the audio device each time before launching the VM?

Thanks for testing. I think the startup problem results from the newly added kernel audio driver. The driver was enabled in both 7.3.3 and 7.4. It now claims the NVIDIA audio device, which shares an IOMMU group with your GPU, preventing GPU passthrough.

As a workaround, add this to the kernel command line for your normal Unraid boot entry, keeping the existing options:

module_blacklist=snd_hda_intel

Then reboot. This prevents the driver from loading, restoring the previous behavior for this audio device without manually unbinding it. Your NVIDIA graphics driver and the VM’s virtual audio remain available.

After reboot, run:

lspci -nnk -s 01:00.1

Confirm that there is no Kernel driver in use line, then start the VM. A Kernel modules: snd_hda_intel line can still appear and is normal.

Please report whether this works after reboot. The earlier shutdown freeze remains a separate, unresolved issue.

21 minutes ago, JorgeB said:

Thanks for testing. I think the startup problem results from the newly added kernel audio driver. The driver was enabled in both 7.3.3 and 7.4. It now claims the NVIDIA audio device, which shares an IOMMU group with your GPU, preventing GPU passthrough.

As a workaround, add this to the kernel command line for your normal Unraid boot entry, keeping the existing options:

module_blacklist=snd_hda_intel

Then reboot. This prevents the driver from loading, restoring the previous behavior for this audio device without manually unbinding it. Your NVIDIA graphics driver and the VM’s virtual audio remain available.

After reboot, run:

lspci -nnk -s 01:00.1

Confirm that there is no Kernel driver in use line, then start the VM. A Kernel modules: snd_hda_intel line can still appear and is normal.

Please report whether this works after reboot. The earlier shutdown freeze remains a separate, unresolved issue.

root@LynxServer:~# lspci -nnk -s 01:00.1

01:00.1 Audio device [0403]: NVIDIA Corporation GB206 High Definition Audio Controller [10de:22eb] (rev a1)

Subsystem: NVIDIA Corporation Device [10de:0000]

Kernel modules: snd_hda_intel

After adding the kernel parameter module_blacklist=snd_hda_intel and rebooting, I ran lspci -nnk -s 01:00.1. There is no "Kernel driver in use" shown, and my Windows 11 VM starts normally.

Problem solved. Thanks for your help, best wishes.

I have another question. When passing through the audio device of my 5060 Ti, shutting down the VM causes the Web UI and VM management commands to freeze. Could you help troubleshoot this issue? Should I create a new separate thread? I recall other users in the forum also reported WebUI freezes with 50-series GPUs when passing through the GPU audio. If this is a hardware limitation with no fix, that is fine. Using the virtual audio card meets my current needs for now. A friend of mine uses Z890 + 5060 Ti with GPU audio passed through. His WebUI remains normal after shutting down the VM.

Thanks for confirming that the boot parameter resolves the startup problem.

Yes, please open a separate thread for the shutdown freeze and link this discussion. We do not yet have enough evidence to call it a hardware limitation. Your friend’s working setup is a useful comparison, though driver versions and GPU configuration can differ.

One possibility is a failure when the NVIDIA driver takes the GPU back after VM shutdown. We need a log from the freeze to distinguish that from a device reset problem.

If you want to investigate further, please include the VM XML with NVIDIA audio passed through and your Unraid and NVIDIA driver versions. Before reproducing it, enable Settings → Syslog Server → Mirror > syslog to flash. After recovery, attach the saved syslog from the flash drive’s logs folder and fresh diagnostics. Disable mirroring afterward. Log capture instructions (https://docs.unraid.net/unraid-os/troubleshooting/diagnostics/capture-diagnostics-and-logs/).

Please also mention whether SSH still responds during the freeze and whether it occurred on 7.3.2.

If virtual audio meets your needs, you can keep that configuration without repeating the freeze test.

6 minutes ago, JorgeB said:

Thanks for confirming that the boot parameter resolves the startup problem.

Yes, please open a separate thread for the shutdown freeze and link this discussion. We do not yet have enough evidence to call it a hardware limitation. Your friend’s working setup is a useful comparison, though driver versions and GPU configuration can differ.

One possibility is a failure when the NVIDIA driver takes the GPU back after VM shutdown. We need a log from the freeze to distinguish that from a device reset problem.

If you want to investigate further, please include the VM XML with NVIDIA audio passed through and your Unraid and NVIDIA driver versions. Before reproducing it, enable Settings → Syslog Server → Mirror > syslog to flash. After recovery, attach the saved syslog from the flash drive’s logs folder and fresh diagnostics. Disable mirroring afterward. Log capture instructions (https://docs.unraid.net/unraid-os/troubleshooting/diagnostics/capture-diagnostics-and-logs/).

Please also mention whether SSH still responds during the freeze and whether it occurred on 7.3.2.

If virtual audio meets your needs, you can keep that configuration without repeating the freeze test.

Thanks,I will create a new thread for the shutdown freeze issue and link this conversation. I will enable Mirror syslog to flash before reproducing the freeze and collect the logs.

Would any if the dockers/host be using the audio where there any messages showing resource was busy when trying to start the VM?

Issue could be that as the audio part of the card is not in the VM it was not released and bound to vfio.