Hyper-V Powering Windows Features

@Craig Really, all sort of apps. I'm constantly observing low rendering speed in my VM, and the more apps I run, the worse UI performance is. Primarily, I use Visual Studio 2019, which seems to significantly affect all other apps rendering (I even reported an issue about that here https://developercommunity.visualstudio.com/content/problem/799943/visual-studio-noticeably-slows-down-the-whole-syst.html). Even when VS performs well, its not enough for many scenarios like working with UI designers. Some debugging experiences are super slow, like opening a Concurrent Stacks view or TPL Tasks list - it can redraw for tens of seconds in some heavy cases. Also, there are many apps which render far not that fast as on the local host, like Outlook, Teams, Skype, Edge (sometimes super slow) and many others. Even windows explorer often renders very slowly. Web UI stands separately, since it can contain some effects, animations etc, which are bad during viewing as well as developing. The same is for developing desktop apps with rich UI and animations. (And there is no point to even mention working with 3D) How can you estimate your "Fluent UI" if it's jut cant be fluent without real GPU. I should note that I have a pretty powerful PC meaning there is always enough physical memory, CPU and SSD throughput for a VM to be superfast, and it is, except for UI. And it makes no difference if I use regular localhost RDP connection, use the Hyper-V console rdp, if I do tweaks to VNet settings like VNet queue etc. Regarding DDA, I'm considering just what currently exists for windows server - injecting an NMVE SSD (I have two and one is dedicated to the VM and both are installed directly in the PCIe through an adapter), and injecting a dedicated GPU card. I consider the latter as an alternative to GPU paravirtualization, though I would prefer to share my host GPU. However, I can imaging many people who would be really happy to be able to use dedicated GPU in their VM for more GPU-intensive tasks, e.g. machine learning and other computations (though I'm not sure that some kind of artist would work with Photoshop or Maya in a VM), This can be not extra expensive if you have an integrated GPU, which can be left to the host, and a discrete GPU, which can be injected into a VM.