Although this post starts a new thread, it's also in response to another thread in the Applications Forum
https://forums.opensuse.org/showthre...GPU-NVidia-GPU
Some references I consider best as of today.
As I described earlier, hardware pass-through(particularly GPU) in virtualization is very new, but is possibly arriving at some stability based on standards like IOMMU and VFIO only within the past half year or so (as of today). So, you're probably looking at this at a good time when everything isn't "The Wild West" when individuals were pushing into unexplored territory, and finally when everyone is generally working on, and improving the same things. But, it's still so early everywhere that there may still be unknown wrinkles that need to be addressed. So, besides what I'll post here, be sure to join communities supporting the specific procedures you follow. And, only follow guides written as recently as possible, preferably within the past 6 months or so.
In any case, this is probably a good time to post something that could be considered a milestone reference for people looking at this for at least the next half year or so.
List of IOMMU supported hardware
https://en.wikipedia.org/wiki/List_o...rting_hardware
Xen
https://wiki.xen.org/wiki/VTd_HowTo
https://wiki.xen.org/wiki/Xen_PCI_Passthrough
KVM
From one of the architects of IOMMU
https://bluehatrecord.wordpress.com/...ith-fedora-22/
Most comprehensive reference
https://wiki.archlinux.org/index.php...rough_via_OVMF
Debian reference, probably all applicable except installation
https://wiki.debian.org/VGAPassthrough
VMware
Requirements
https://kb.vmware.com/selfservice/mi...rnalId=2142307
Configuration
http://www.dell.com/support/article/...hrough?lang=EN
Virtualbox
As usual, VBox is a bit behind the others without a clear guide for enabling GPU pass-through. The following link to the VBox Advanced Configuration documentation, specifically PCI passthrough in general and with a section on passing through a webcam
https://www.virtualbox.org/manual/ch...pcipassthrough
As I noted in my prior post,
If you consider a non-virtualization technology like docker or LXC, you likely wouldn't have to deal with virtualized devices and pass-through at all because Linux containers always have direct access to hardware devices. I haven't researched this approach but I do suspect you would only need to assign resources using tried and true common methods that have been around for many years.
Particularly for this technology, also search for GPU topics related to your chosen virtualization in this openSUSE Technical Help -Virtualization forum for related information that might cover developments and issues not mentioned in the above references.
TSU