if special drivers are needed for the hardware, these will need to be installed in the VM
guest OS, and you will no longer be able to live-migrate the VM.
When mapping an NVMe storage device to a VM using DDA, no additional
configuration is required. However, if you are mapping a graphics device that tends to
have a lot of memory, memory-mapped I/O space will not be available for the guest
OS to map and use the memory of the graphical device. You will need to reconfigure
the VM. The exact amount of memory that needs to be mapped will depend on the
graphics device, and it can be seen by examining the Resources tab of the device in
Device Manager, as shown in Figure 2.21. You can modify the low and high memory
space for VMs by using the - LowMemoryMappedIoSpace and -HighMemoryMappedIoSpace
parameters of the Set-VM cmdlet. This is explained in detail in this article:
Figure 2.21 Memory resources for a graphical device
https://blogs.technet.microsoft.com/virtualization/2015/11/23/
discrete-device-assignment-gpus/
The article also points out that you should enable the guest to control the caching by
setting -GuestControlledCacheTypes $true via Set-VM. All of these configurations
require the VM to be shut down.