Framework Laptop 13 pro | Help needed with the choice of hardware

User Support General
1

I am planning to buy a Framework Laptop 13 Pro and install Qubes on it, and I have a few questions regarding this:

  1. I am choosing a CPU, I am mostly interested in ‘Intel Ultra X7 358H’ and ‘AMD Ryzen AI 9 HX 370’. According to the benchmarks, the Ryzen AI is going to perform better, and I personally prefer AMD over Intel, but as far as I know, Intel processors are generally more recommended for Qubes, due to better support and faster security updates. What should I take into account when making this choice?

  2. I couldn’t find a lot of information on this, but if I go with Intel, I would like to utilize SR-IOV to “cut” my iGPU into smaller ones, and assign these virtual GPUs to VMs. As far as my research goes, from this page the SR-IOV is unsupported on ‘Panther Lake’ family, but is there any chance it will be in the future?

  3. Is 32GB of RAM enough for comfortable multitasking on Qubes? (eg. Browser opened, text editor opened, music player opened, messenger opened, something else is opened)

Thank you in advance for the answers.

2

I haven’t tried any Intel Panther Lake CPU yet (like in the above Framework), but I tried Intel Lunar Lake (one generation older). And there are issues with the new “xe” GPU driver under Xen: Graphical glitches (not refreshing?) with Linux’s xe driver + Xen 4.19

This should eventually be fixed, and also you’ll wait quite some time for that Framework laptop to arrive (looks like buying now will get you one in September), but right now this one specifically may be problematic.
As for SR-IOV, oh, they retracted the support from PTL too :frowning: Something similar happened to LNL before - at last minute the commit enabling support there was dropped. But for PTL, the support seems to be included in Linux (Making sure you're not a bot!) since quite some time. But, here also I didn’t got much success in practice (on LNL, I don’t have PTL, yet).
I have seen people reporting successes with i915 + SR-IOV patches (there is a DKMS module that retrofits old Intel patches, before they stopped working on official SR-IOV support for i915 driver). But with this approach, you’ll face another issue - i915 driver doesn’t support LNL or newer anymore…

So, the situation on the Intel GPU side doesn’t look great right now, but there is hope :slight_smile:

As for AMD, yes, they are usually slower at delivering microcode updates for desktop/mobile CPUs. And also, I don’t think they even plan supporting virtualization with integrated GPU (or desktop models either for that matter). So, from the feature/security point of view, Intel seems like a better choice. But, right now, with the switch from i915 driver to xe driver, there seems to be a couple of reliability issues.

As for memory amount, I’d say yes. My daily driver has 32GB and it’s perfectly fine.

3

Yes. Running a lot of qubes with different tasks on 32 GB simultaneously.

But…if you plan a dedicated HVM for local AI via ollama, go for more. This eats up 16 GB alone on my system.

Same with Windows HVM.