Depends dude. Have you actually run 4k on your TV at regular viewing distance?
My living room pc is hooked up to a 55" 4k TV.
I run it in 1080p. Why? Frame rate, and Ubuntu desktop in 4k is fucking impossible to read at 6 feet away or so. If it wasn’t ever in the desktop 4k would be fine, if slower… but…
I’m normally less than two feet away from the TV. Haha
Small text on such a big screen means I have a lot of real estate for multitasking. It’s probably close to how someone would use two ultrawides stacked vertically.
Ah fair enough. At that viewing distance, sure. 
Another perspective: since you are looking to experiment and willing to invest now and again when some more experience is gained, I propose to setup a system using Proxmox VE. https://www.proxmox.com/en/proxmox-ve
It is a debian based Virtualizing environment, free to start but you can get paid support if you need to.
Fun attributes: You can run it on a headless computer and administer it using a web interface, then setup containers and VMs. The LXC containers can also be CentOS or other linux distro’s, but ytou can als use KVM to run Windows and other VMs.
Another interesting property is that it can run of ZFS, and that you can build VM clusters. So when you build your second machine you can just tell the first machine to move a VM to the new machine. Or build a High Availbility cluster! ZFS has interesting snapshot features. Or you can use CEPH to build redundant networked storage…
I have an experimental instance running at home on a simple headless box. One caveat: any system meant to run VMs or ZFS needs lots of RAM. If you need to rely on such a system, consider building with ECC RAM, (using used server hardware for lower cost), with redundant power supplies.
I really like the fact that the base VM machine has low overhead, but provides a lot of possibilities. And I learned to really appreciate PuTTY and WinSCP for easy access to consoles and files from my normal Windows based family computer. Generate some keys, put thim in the right authorized_keys file, and BAM, easy to open up a console to administer stuff. Also editing some config file is just a couple of clicks in WinSCP.
This seems like a useful tool, and I’ll check it out. I’m juggling a few projects right now and can’t focus enough to digest your post, but I’ll give it another read when I can settle down.
Status update:
I learned that Facebook Marketplace is a decent source for certain kinds of used gear, I’m confident I can find a used RX 580 for under, or just over $100 USD. Haven’t shopped for other cards yet, but that’s promising. There’s also randomly a Dan-a4 for sale in the south of Taiwan. But they want more than $300 USD for it…
I took the suggestion of “just virtualizing Linux in Windows” in the meantime, and it is clunky to me. I decided to give Elementary OS a shot since it’s supposed to be friendly to computer illiterate folks and there’s already little frustrating quirks that left me needing to take breaks… most of them were related to actually setting up VirtualBox’s integration, but also not being able to right click on the desktop and put things there feels really stone-age. The “super” key bringing up a shortcut menu rather than doing something like bringing up an application search is a minor frustration, too.
Still, I set up Python and have made it through to Chapter 4 of Automate the Boring Stuff. I have sat down to do Chapter 4 a few times, but can’t seem to focus on actually reading the material and keep catching myself just diving down to where it asks me to write a program. I know that’s bad form, so I set it down when that happens.
That’s where I’m at right now.
I’ve been trying to decide on the right motherboard. I like the DTX board because of how obnoxious it is, but it’s almost double the cost of simpler ITX boards. There’s one ITX AM4 board that supports ThunderBolt 3, which is really cool. I can’t decide if that’s more valuable than having two M.2 slots… I think the right choice is to go for ThunderBolt 3 and just make a NVME SSD my “nice flash drive,” and I can just leave it with my desktop most of the time.
Well… I haven’t had any support or driver issues and any screen tearing that occurred in KDE on my laptop was resolved by switching to Wayland. With KDE, that brings its own bag of issues, but it’s just fine to do so with GNOME.
Yes, I did read the replies to the quoted comments, but I still wanted to give my two cents on the matter.
Well… Regarding integration, I’ve had better luck using VMware Workstation Pro or VMware Player (I know, heresy), but VirtualBox should be getting a boost in the next few updates.
Regarding placing things on the desktop, I hate it when people do that. I respect other people’s opinions and preferences (KDE supports desktop icons if you really must have that and the super/meta key behavior is what you would expect coming form Windows), but seeing people place icons on the desktop annoys me, especially when they fill the entire desktop up, as if no other folder exists on the drive. And how is it stone-age if it is a relatively recent development in desktop environments?
Something you might want to try is Unraid. It could check some of the boxes that you a looking for and has great community behind it. This is certainly not what I think most people would suggest but it is a great way to get you feet wet with Linux and hardware pass through and friendly GUI go a long way.
You said you wanted a Windows VM that was close to native performance. Well, with GPU pass through, you could pass through your graphics card (probably a USB controller too since Unraid boots from USB) and then you would have a very close baremetal experience.
You mentioned peeling back the onion. With Unraid you could you use the VM manager to spin up a Linux daily driver ( you could vnc or also pass through a secondary GPU) quickly determine what DE works for you. The thing that I like about this is that you could try out a lot of different flavors while still having a rock-solid everyday OS.
I just like to use my desktop as a “cutting board” type space a lot of the time. I’ll make a file or folder there when I start working on or collecting things, and then I put it in a place it makes sense when I put the project down.
Having things on the desktop makes more sense when you primarily drive your computer through the desktop environment and with a mouse. Which, I get a lot of Linux users either don’t do that or use the keyboard/terminal a lot more when they can.
But either way, this is just a matter of preference!
I’m interested, but I’m not sure if it’s the right solution. My main hurdle is that I’ll only have one GPU, and if I have a Ryzen CPU, it’s very likely I won’t even have a “integrated” GPU in my CPU.
My hope was that I would be able to boot into Linux, using my GPU (anticipating an RX 580 still), and still be able to pass the GPU through to a virtual machine.
I know there are sometimes issues with “sharing” consumer-grade GPUs, and I could live with “unassigning” the GPU from the Linux host while my Windows VM is running, if that is possible.
Eh, I generally use a combination of the terminal and the file explorer. To each their own.
As far as desktops which support desktop icons go, you have KDE, XFCE, LXQt and a few others which I either haven’t heard of, or which escape me at the moment. From what I know, Pantheon and GNOME do not support desktop icons, though Canonical have a GNOME extension set up on Ubuntu to enable desktop icons, though I haven’t tried Ubuntu Eoan Ermine yet.
With that said, assuming you’re not trying to run two Unraid VMs at the same time with passthrough, while it wouldn’t be my first recommendation, Lonercoin’s suggestion is not bad at all. The two points above were addressing a Linux-host-Windows-guest situation by the way.
By the way, the driver blacklisting occurs at boot, so, as far as I know, you can forget about having the card go between the host and the guest without a reboot or two.