My daily driver already runs Qubes OS, and I have a separate machine that I’m considering using for creative work—especially video editing. I’m deciding whether to install a conventional, monolithic operating system on that machine or use Qubes OS and pass a dedicated GPU through to a qube. I feel safer using Qubes, so that has my preference.
Would GPU passthrough let me do video editing comfortably, including GPU-accelerated effects and rendering? Would I still get meaningful security or compartmentalization benefits from Qubes in that setup? I’d appreciate advice from people who’ve tried it, especially about performance, software compatibility, and the setup’s reliability.
I have past experience in video editing but not in Qubes.
GPU acceleration is not officially supported, and IIUC these passthrough setups are somewhat hacky, so the safety you feel now may not be the same with additional graphics hardware that boots and runs alongside your Qubes certified PC.
I would approach it this way:
Is security top priority?
Will video editing be done for making money?
Can I afford 2 separate systems?
Any “Yes” is a vote for 2 separate systems.
For a start, you can try your video editing software to see if it will work in a qube at all.
Thank you for responding Yes, security is top priority but I also need to make a living an I need to be able to do creative work for that. I really love Qubes, but sadly, it’s a bad fit for us creatives who need high-end graphics tools, which often require a lot of RAM and GPU power .
I will do what you say and try a pass-trough in the other machine running Qubes. Worst case scenario I will move to a monolith and spend some extra money just getting another network and doing all creative work on a completely different network - so I’m sure the one I’m on now remains safe. And when I have some extra money to spend I can consider an external gpu for this machine.
If anyone with a working external gpu feels like making a comment about their experiences please do.
Testing your editor in a qube before committing to passthrough is the right first step, as qubist suggested, since that will tell you quickly whether the software runs at all without a GPU. On the security side, once a GPU and its drivers are attached to a qube, that qube effectively trusts the hardware, so the compartmentalization is weaker than for your other qubes. A separate physical machine on its own network, as you’re considering, keeps the creative work from touching the Qubes hardware at all, which sounds like the more reliable path if editing is how you earn money.
The point about the passthrough qube trusting the GPU once drivers are attached is the part worth weighing most heavily. Since editing is how you earn money, a failure there costs you directly, while a separate monolith on its own network keeps that risk off the Qubes machine entirely. Trying the editor in a plain qube first is cheap and tells you whether hardware acceleration is truly required. If it is, I’d lean toward the second machine and treat an eGPU as a later convenience rather than the foundation.
.