Anyhow I think the EFI work will just exacerbate the issue (now it will run on Apple laptops with discrete graphics?), and am willing to help however I can (including code work)
Is that documentation up-to-date? It refers to Fedora 18 and kernels in
the 2.6.x series!
One of the Qubes devs could provide a definitive answer, but I suspect this is the reason for prioritising EFI.
I see no point of using a dedicated GPU in Qubes if there is an internal one. (Except for additional video output that is not controlled by the internal GPU.)
I've successfully installed Bumblebee on Qubes 3.0 several months ago and it still works fine. I don't have Nvidia proprietary drivers and Nouveau don't seem to be used. There are my steps summarized: https://groups.google.com/d/msg/qubes-users/s2oOT1nTFkw/GflmP2FmFAAJ
Performance
ability to run multiple external monitors, ports not available to internal GPU as you say.
I've successfully installed Bumblebee on Qubes 3.0 several months ago and it still works fine. I don't have Nvidia proprietary drivers and Nouveau don't seem to be used. There are my steps summarized: https://groups.google.com/d/msg/qubes-users/s2oOT1nTFkw/GflmP2FmFAAJ
I have an AMD chip
Regardless the point is that from what I'm seeing the GPU is the most difficult issue with installing and using Qubes at the moment.
I can't imagine any performance advantage over integrated GPUs when using with Qubes. That is, Qubes uses GPUs just in dom0, where you can get WM acceleration. But integrated GPUs are good enough there.
--
You received this message because you are subscribed to a topic in the Google Groups "qubes-users" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/qubes-users/NnO9NFdpGbw/unsubscribe.
To unsubscribe from this group and all its topics, send an email to qubes-users...@googlegroups.com.
To post to this group, send email to qubes...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/qubes-users/b1be5ecb-ced0-4af2-85b8-df24cc15c0bc%40googlegroups.com.
out of the box Qubes seems to be made for low end machines.
First is memory, with 32GB Qubes actually suffers, so I need to limit memory in the VM's.
Two is having 4/8 cores, so limiting vCPU's fixes that (by default they all get 8 cores and then suffer from contention).
Final thing is that the graphics still seems to stutter too much, I'm still digging into it (might not be GPU related but inter-VM communication or something)
On Thursday, December 10, 2015 at 12:26:48 PM UTC+1, Cube Mammal wrote:out of the box Qubes seems to be made for low end machines.I don't think so. Qubes is rather memory-hungry, SSD is strongly recommended (I've tried running it fully on HDD, so I can confirm that having at least some parts of Qubes on HDD is a good idea…) and various features of CPU (most notably IOMMU (VT-d)) are needed for some features.
First is memory, with 32GB Qubes actually suffers, so I need to limit memory in the VM's.Well, some issue related to this seems to be addressed by limiting dom0 RAM to 4GiB. Not sure if this is exactly the your issue.
BTW, I haven't seen such problems with 16 GiB RAM, which is far from being lowend.
What workload type do you use? Where do you feel the stuttering?
Oh I agree with you but felt that 3.0 didn't have the settings for a high end machine. Out of the box 3.1 does however, it comes with a 4GiB limit on DOM0
I notice Qubes does better if you limit sys-* VPN's to a vCPU of 1
and give 2 to the AppVM's.
When using multiple monitors the graphics slows down and stutters, eventually crashing, both on 3.0 and 3.1
With the integrated graphics and the existing panel it runs beautifully, though like I say running the exact same machine with Ubuntu or something (AND three external monitors with discrete graphics) has no issues. Same issue on the same laptop with Nvidia graphics. Qubes just doesn't work well with discrete graphics, at least in my tests.