I have a bug(?). In Qubes 4.3, if I open a video in a qube using mpv, everything works fine. But if I then open a second video in the same qube using mpv, the qube immediately freezes (i.e. slows down to the point that I cannot interact with it or even with dom0). Using htop I have confirmed that this is not happening because the qube runs out of memory or cpu power. I did not experience this problem in Qubes 4.2.4. I’m using an AMD Ryzen 9.
I’m just using mpv as an example. The same thing happens if I open any two programs that consume a reasonable amount of memory / cpu. But it doesn’t happen if I open, say, mpv and mousepad at the same time.
Are you sure about this though? When a VM is extremely overloaded the diagnostic tools may no longer be reporting accurately as they too are being strangled. Perhaps you could check journalctl -b in the VM (once things settle, or after a reboot) to see if there is better retrospective info there.
Look at the tips here about improving mpv performance on Qubes:
Also, you can bump up the vcpu value of the Qube so it gets more CPU priority. That could make a big difference in the scenario you describe.
I set up the xfce panel widget that displays cpu usage (xfce-cpugraph-plugin). When I opened two programs in a qube (e.g. two mpv videos, or an mpv video and a video game, &c.) a single dom0 cpu process maxed out while the other processes remained idle. My computer became so slow that it was almost unusable. I do not think that this is a qubes problem. There is a xen bug in which, if you run a certain version of xen with certain boot-time flags, some processes are not multithreaded. This problem may be unique to my architecture.
I fixed my problem by switching from X11 to wayland in dom0. When I was running xfce there was a single /usr/bin/X process that maxed out a single cpu whenever anything with graphics was running on both of my monitors simultaneously (i.e. not just mpv). In htop in dom0 /usr/bin/X was mostly red, which I think means it was doing a lot of kernel-level memory stuff.
I did not previously have this problem in qubes 4.2.4. Either the people maintaining x11 changed something between these two versions of qubes or x11 was compiled with different compiler time options or something else.
When Qubes 4.3.0 was first released, there were a lot of threads in which people were saying that it was a lot slower than 4.2.4. Maybe they were having the same issue.