I've been using VMs hosted in unRAID from a variety of different clients over WAN and LAN and have been more than happy with the performance for most things.
Chromebook using Chrome Remote Desktop again absolutely fine.
MK808b+ SOC running Android using Microsoft RD app from App store has been fine (port forwarding & IP info required)
VNC using RealVNC again absolutely fine (port forwarding & IP info required)
This across a mixed range of VMs and Clients, Linux and Windows. Pi would have zero problems imo. The MK8080b+ SOC is similar hardware. Using Remmima or Chrome Remote Desktop to control the VMs on the server.
However, this was all for standard stuff - coding, some browsing, nothing major. Youtube not a great experience with vGPU.
Now for the tricky part. I'm hoping to do a little more heavy lifting on one of the VMs in the unRAID server and do some 3D modelling work. My server currently doesn't have a second GPU (runs headless) and so all VMs are using emulated GPUs.
Will passing through a dedicated GPU to that (Windows) VM that needs to do some of the heavy lifting (bearing in mind that it will still be accessed over one of the protocols above) improve performance? Just trying to figure out whether or not I really need to add a GPU for this task.
On the same subject, given that RemoteFX requires Hyper-V, am I right in thinking that there is no way for unRAID KVM to harness this? (Even if I host the VM on a hosted Windows Server within KVM?)