Librem 13 freezes on very high load

If I’m running two vms (Windows 10 and a Linux box) and I know that the performance drops significantly because of that but I don’t expect, that it freezes.

After a reboot it complains about high temperature and that it throttles the cpu (which should be done in the first place, BEFORE it freezes and I have to turn off the computer) so it’s something about temperature I suppose.

What can be done about this? I’ve opened the Librem, it’s completely clean, so it’s not dust/dirt…

Temps ain’t gonna solve everything, but if you need to get better temps try using a non conductive(electrically) thermal paste like arctic mx4. This problem could be caused by cache misses (speculative execution failures), memory bandwidth issues, the virtual machine itself, etc.

My guess is you are using boxes to virtualize, that uses libvirt which isn’t a KVM (kernel virtual machine). meaning there is a high performance overhead. Try using QEMU in KVM mode. Give each vm 1 thread / 1 logical core, that way you have 2 threads (1 physical core) to the host OS. That way execution is limited to 50% of the entire CPU keeping temps cooler. Also boxes allocates 50% of the CPU per virtual machine by default I think.

1 Like

I’m running libvirt/kvm and the vm has two logical cores assigned, the linux vm one core and I’ve changed the netstructure from the win 10 vm to 1 socket, two cores and four threads, because it feels and is pretty laggy…
That seems a little bit to much if a Linux vm is running at the same time, which just uses one logical core.

But still, does Debian not issue some kind of performance warning to the user, if it’s overloaded? Can I configure something like this?

Thanks for your help btw!

How many logical cores does the windows 10 machine have? And how many for the linux machine? How many cores total in your system? Libvirt virtualization and KVM are different. KVM has like 2% overhead, libvirt has more than that ofc. What software are you using? If you wanna check load on cpu go into the terminal and type

“sudo apt install htop”

wanna see temps

sudo apt install lm-sensors
sudo sensors

I’ll report back when I’m back at my computer :slight_smile:

The Performance Co-Pilot looks like a very good tool to diagnose this issue as well, we use it internally.


That looks like a great tool! I didn’t even know it existed :slight_smile:

1 Like

As I said above I’m using two logical cores for the win vm and one for the linux vm. So in theory there should be one free logical core for the host system available?

I’m using the “virtual machine manager” running with libvirt and changed the netstructure to the default one (which are two sockets, one core and one thread) now to check if this is the real reason for the freezes.

running with these settings seem to be fine but didn’t check the temps, but I’ll do if it happens once more - installed lm-sensors, htop and pcp to monitor it in the future.

1 Like

but do you only have 4 logical cores is what i wanted to know.

I think yes, two physical cores plus two hyperthreading cores

yeah you should never give up more than 2 logical cores / 1 physical cores. You need at least 1 physical core at all times

1 Like