Yeah I’m pretty sure with some simple virtual network bridges and routing rules a 4-to-6 converter is plenty possible without touching the actual processes.

Based, if I can, I will edit the original link to use piped

Well at least for Nvidia, vGPU is a fully Enterprise tech for accomplishing splitting a GPU between a host and VMs. It also just so happens to work on all 20> series cards if you patch the driver :}

You can not just pass through, but share any¹ gpu you like using HyperV. Yes it’s Win10Pro, but there are pleeeeenty of ways to get it enabled/installed/supported on Home as well. Though if you have an Nvidia card 20 series or older, and you’re willing to dive into linux as a dual boot, I’d say qemu/virt-manager is a pretty mainstream VM solution, and vGPU is also a good tech for the same purpose.

₁ I’m not actually sure what the limits on hyperv are but it seems fairly robust. Don’t quote me on it lol

Specifically to set up a system like the one I’ve finally got. A hypervisor that retains its own display capabilities, while being able to share the full GPU (though only set portions of VRAM) freely and on demand, with a VM. We can shit on Nvidia all we’d like for a lot of stuff, but vGPU works really well, even on cards it isn’t meant to support LOL

Between the discord devs outright refusing to do any kind of sound capture for linux screen sharing for several years, many updates requiring you to manually download and install a fresh tarball instead of being automatically applied like on windows, and refusal to maintain any kind of package on most repositories, I’d say it doesn’t properly support linux. I do like Nobara’s version of it, whatever they’ve done (I haven’t looked much into it but it definitely seems custom to Nobara, or at least Red hat) though.

And speaking purely from personal experience with no real way to verify statistically (so take this with a rather large grain of salt), there are a LOT of CS or CE major types that would love to switch to it, but will be faced with random tools that they need like microchip studio, or some particular CAD software, just not working at all. For those that I’ve talked with at any length, if they could spin up a VM that effectively fully works as if it’s bare metal, including proper display out that matches the monitor, whenever they need those few specific things, they would switch. They may not be many compared to 100% of global internet users, but they could certainly make up for 2%.

That said I mostly agree that devices like the steam deck are where linux is gonna grow, I just think that it would be going faster if more devs were able to daily drive it and care more about it, instead of having to be stuck reliant on Windows

That’s where you’d be wrong! I’m running it on my 1080ti right now. It can be hacked into working on just about any Nvidia card that’s recent enough to want to use it. A bit of a community has ended up growing around a group that makes patches for the official vGPU drivers, along with merge scripts, to give the hypervisor the ability to retain regular function (accelerated display out through the DP/HDMIs), while also fooling the vGPU part of the driver into thinking the random consumer card is supported. Unfortunately locked down on 30 series and newer :(, but it’s still a VERY cool use for a card like the 1080ti that has become VERY cheap

Close, but that’s not what I mean. I mean SEAMLESS sharing, not playing tennis with it. Is it really so poorly known? Should I write up a little introduction to the parts of it that I’m familiar with?

As someone who has FINALLY made the switch to contribute to that (all it took was several whole days of the summer devoted to constructing a Void environment from scratch to host VMs with GPU sharing), I gotta say the SteamDeck must be putting up gooood numbers, cause there’s no way enough people full-time switched.