I am running kde plasma 6 on fedora and when running on my rtx 3080 i am having random lag spikes on my desktop, to the point opening system settings occurs lag spikes or occasionally using the terminal.
using nvtop my gpu has an average idle usage of 30% but sometimes it randomly spikes to 80-90%. For my nvidia drivers i am currently using the discrete drivers version 580.
also for nvtop the highest usage pids
are /usr/bin/plasmashell --no-respawn - gpu 0% (this switches between wayland and the plasmashell)
second is usr/bin/kwin_wayland --wayland-fd 7 --socket wayland-0 --xwayland-fd 9 --xwayland-display : 0 --xwayland-xauthority /run/user/1000/xauth_tFicTx --wayland -
I also noticed that this process randomly spikes to 100% CPU usage
any idea to fix this? completely new to Linux so any help is much appreciated!
Not normal - hereās my 2080 running KDE (clipped from the top of the nvtop window if you want to compare. Should be idling in low single figures and spiking to 100% use should not happen unless youāre actively using it. In fact spikes period should not happen without you having a clear indication of why.
Device 0 [NVIDIA GeForce RTX 2080] PCIe GEN 1@16x RX: 390.0 KiB/s TX: 3.909 MiB/s
GPU 300MHz MEM 405MHz TEMP 45Ā°C FAN 0% POW 23 / 215 W
GPU[|| 5%] MEM[||||| 1.216Gi/8.000Gi]
Post the output of inxi -Fzxx so we get a good look at the kit youāre rurnning on, make sure which drivers are loaded and so forth.
You could also run an fpaste --sysinfo --printonly command and paste the output here as pre-formatted text (the </> button above). If the output is too long then remove the --printonly parameter and itāll send it to pastebin and give you the URL back.
Nothing looks too amiss in the fpaste output, apart from the pretty high load average.
* Load average (uptime):
16:16:10 up 3:28, 2 users, load average: 6.33, 7.01, 8.14
and the relatively high CPU temperature for a system which Iād expect to be idling.
System Temperatures: cpu: 64.9 C mobo: 49.2 C
Are you aware of what is currently running on this machine and itās all as youād expect? I ask this because Iād expect to see load averages under 1 as a rule, unless someone really does have a large package build going on or a kernel recompile and so forth. Maybe you really do have a load of background jobs currently active - I donāt know your use case.
You can see what is actually running or uninterruptible with something like ps -eLo pid,tid,stat,comm,%cpu --sort=-%cpu | awk '$3 ~ /R|D/'. Feel free to post that output as pre-formatted text.
i got this as a result - luis@fedora:~$ ps -eLo pid,tid,stat,comm,%cpu --sort=-%cpu | awk ā$3 ~ /R|D/ā
41406 41406 R+ ps 200
678 678 D kworker/u50:10+ 1.8
luis@fedora:~$
well i was using kvantum for themes and a few global themes but that is pretty much it.
But I think i was able to narrow down the issue.
I reinstalled my OS a few times to check what is the determining factor and found this.. on a brand new install doing sudo dnf update seems to automatically install the gpu drivers but with some issues.
im presuming drivers are installed since on system settings its naming my gpu with the discrete label on it, and it has cuda installed after doing the update.
and after rebooting it, it has the random spike usages again. so i presume the issue must be coming from the drivers itself potentially
sudo dnf update on a new system probably updates the nvidia-gpu-firmware package, but not the actual Nvidia drivers. Those wonāt be installed unless you explicitly install them.
If you want to see for sure what driver is in use, then do fpaste --printonly --videoinfo
In the output, look for āKernel driver in useā under your video card - this should state either nvidia or nouveau. In your earlier post it was nvidia, but if you do it on a newly installed system you should see nouveau until you install the drivers and restart.
Just had to force reboot my computer due to plasma using all my vram. I have a low end card but KDE has been using almost all my vram since the latest updates.
I normally have two or three sessions open but since the latest Plasma updates, I have a hard time with two sessions open. I suspect some setting in my one session is causing weird issues but kscreenlocker is using over 200 meg for a locked session. Firefox with hundreds of tabs open is using less vram.
If you replicate this situation, can you use nvtop to kill the impacted session (F9 is the shortcut key) - does that kick things back into life and release the memory consumed or does it appear to be leaked, never to be seen again until a reboot?
For reference, hereās my RTX2080, using Plasma 6.5.1 and the open nvidia drivers. Not sure itāll help much, but it may give you a reference point for comparison.