Performance regression and massive FPS drop from 60 FPS to 30 FPS in games using Plasma 6.7.0/6.7.1

Help
, , ,
22

Has anyone opened htop when games are showing low performance and high overall CPU utilization to see if kconnectd is using 100% of one core yet? On 6 and even 8 core CPU’s, one core maxed out at 100% will substantially affect performance, and kconnectd has been hogging a CPU core randomly since the update to 6.7.0.

23

kdeconnectd I guess?

24

Semantics.

Here’s a screenshot, the only reason my FPS doesn’t take a huge hit is because I’ve got another 13C/26T’s to take up the slack:

100_2%
100_2%5120×1440 1.29 MB

25

Yeah I saw that issue as well KDE Connect randomlly locking one core at 100% since update to KDE Neon 6.7

Htop on my end (for GTA IV at least) peaks at ~40% for a single core, KDE connect wasn’t hogging the CPU at the time I recoded the last video.

Not sure what else to check/do, I wish dependencies on Linux were much easier to manage so I least I can downgrade/upgrade single components and see what regressed between v6.6.4 and v6.7.x and provide useful report to devs, but sadly that’s not the case and building KDE components is really difficult for first timers.

Thanks everyone by the way for all the suggestions, for the sake of my sanity I ended up gutting out all KDE/Plasma packges entirely and installed COSMIC desktop. Everything just works, all OpenGL/Vulkan games either using wayland composition or X11 (XWayland) just work as expected without a single FPS hitch.

For devs, I kept a working btrfs snapshot of the entire system with Plasma v6.7.2, if at any point in time you want me to test something or provide logs I can easily switch the mount subvol and I’ll gladly help as much as I can.