hey,
sorry if this is a duplicate topic but i didnt found a real solution / explaination for my problem.
sometimes im working on my multimedia qube on my server (bad habbit i know - i should do here multimedia stuff instead of administer a server), and if im working im going to be a tab hoarder, because, you know if you are in a forum, then someone is posting a link, then you find another link,… because all things are related.
anyhow, spotify is one of these programs which is running in the background all the time on this vm, the browser im using is brave.
mostly if i got more than 10 tabs opened my cpu usage would walk up to 100% while other vms are still fine.
is it possible to passthrough bit more cpu power to multimedia? and why is (guess) brave using that much?
guess this cant be related to the hardware acceleration, can it?
Try setting your CPU to performance mode:
In dom0: xenpm set-scaling-governor performance
Put it back to normal with: xenpm set-scaling-governor ondemand
These do not persist a reboot.
tried your suggestion out now because i rant again in this behavior / still having it because i got a longer session in multimedia.
if i try your command with performance ill get back:
[CPU(number)] failed to set governor name (22 - Invalid argument)
im pretty sure there is no typo because i googled your command before to lookup if its correctly spelled
let’s see what your CPU is capable of with:
xenpm get-cpufreq-para
cpu.log (3.4 KB)
as you can see some cpus failed to get cpufreq parameter
i got a intel i7-10850H in it
im on a lenovo thinkpad p15v
mainboard is new because it got changed ~2 months ago… so cpu and mainboard should be fine
update:
i were so smart and run top while i was watching youtube and do some things, i found out, that isnt brave which is using ~90% of cpu usage, its kswapd0.
have to find out what kswapd0 is and why its overloaded sometimes.
maybe you got a guess before i would find it out? anyhow, its for me slowly too late to continue troubleshooting.
little update here:
i just waited some time to see if this issue would come up again since im using fedora 37
seems like this issue is gone with fedora 37 upgrade