I have recently downsized my unRAID server to an i5 8400 6 core CPU. However I noticed that when running a Windows 10 VM, the CPU usage seems quite high. If I use "top" in an unRAID ssh session it shows qemu-system-x86 as the highest consumption. On the dashboard I can see one of the cores running at 45% and the others moving around 5-15% up and down. The actual VM is doing nothing and task manager in Windows 10 shows 99% idle. My previous 14 core Xeon never used to do this at all, if the WIndows VM was idle then so was the server largely.

 

Any ideas?

I also get a fair amount of CPU usage when all I have is a single VM suspended? If I actually stop it properly then the CPU drops off. I thought suspended meant basically hibernate?

 

UPDATE: okay read the redhat docs on this and suspend just suspends in RAM aso I could see how some CPU is used if qemu is still running in the background. I need to set my scripts to do a managedsave instead. I still have the initial issue though. Given that CPU usage is still high when suspended I think I must have a lot of ambient CPU use by qemu itself for some reason.

Edited July 11, 20196 yr by planetwilson

Okay did a bit of digging and found this thread:-

 

https://askubuntu.com/questions/1033985/kvm-high-host-cpu-load-after-upgrading-vm-to-windows-10-1803

 

<clock offset='localtime'>
  <timer name='rtc' tickpolicy='catchup'/>
  <timer name='pit' tickpolicy='delay'/>
  <timer name='hpet' present='yes'/>
  <timer name='hypervclock' present='yes'/>
</clock>

All much better now, apparently the hpet line is important

Edited July 11, 20196 yr by planetwilson

One of the reasons for high CPU usage is that Windows seems to throw a lot of interrupts.  I've had mixed success with enabling HPET - sometimes it works, and sometimes it doesn't.

 

What I have found that works all the time is disabling the unused USB Root Hubs.  CPU usage will drop to nearly nothing at idle when I do that.