** Description **
On my laptop, I noticed that after a few hours of working, plasmashell would be using lots of CPU. Restarting plasmashell returns it to using only 0-1% CPU, but then the CPU usage slowly increases again over time. After some investigation, I discovered the following:

1. The rise in plasmashell's CPU usage is caused by using animated widgets on the desktop (CPU/network monitors, clock, etc.).
2. This bug affects my laptop (open-source radeon driver) but not my desktop (NVIDIA proprietary driver). See system details below.


** Steps to reproduce **
1. Add the "Analog Clock" widget to the desktop and resize it to fill the entire screen.
2. Enable "Show seconds hand" in the widget settings. This causes the clock to have frequent animation which is necessary to trigger this bug. On both of my systems, plasmashell uses only 0-1% CPU at this point.
2. Let it sit overnight. Don't lock the screen or allow the PC to suspend, but monitor blanking/power saving seems fine (i.e., it doesn't interfere with the progression of this bug).

Result: After about 7 hours, plasmashell CPU usage has risen to 12-18% on my laptop. The clock's animations are now jerky (the little bounce when the seconds hand moves is no longer perceptible). If I attempt to rubber-band-select on my desktop, the animation is very choppy. Restarting plasmashell makes everything smooth again, and plasmashell's CPU usage drops back down to 0-1%.

My desktop, on the other hand, does not appear to be affected by this bug. After 7 hours, the animations are still smooth and plasmashell's CPU usage is still between 0-1%. Both systems are running the same OS (KDE neon) and plasma version (5.10.0).


Here are the details for each system:

** My Laptop (the affected system): **
OS: KDE Neon 5.10.0 64-bit (Plasma Desktop 5.10.0, KDE Frameworks 5.34.0, Qt 5.7.1)
PC: HP Pavilion m6-1035dx
CPU/GPU: AMD A10-4600M APU with Radeon HD 7660G Graphics (using xorg radeon driver)
RAM: 6GB DDR3 800 MHz
Linux Kernel: 4.8.0.54-generic
X-Server: 11.0
Renderer: Gallium 0.4 on AMD ARUBA (DRM 2.46.0 / 4.8.0-54-generic, LLVM 4.0.0)
OpenGL version: 3.0 Mesa 17.1.0

/etc/X11/xorg.conf:

  Section "Device"
      Identifier "Card0"
      Driver "radeon"
      Option "AccelMethod" "glamor"
      Option "TearFree" "on"
      Option "DRI3" "1"
  EndSection

Kwin compositor settings:
  Rendering backend: OpenGL 2.0
  Tearing prevention ("vsync"): Never


** My Desktop (not affected by the bug): **
OS: KDE Neon 5.10.0 64-bit (Plasma Desktop 5.10.0, KDE Frameworks 5.34.0, Qt 5.7.1)
Motherboard: ASRock X58 Extreme3 (Intel X58 chipset)
CPU: Intel Core i7 930 (2.8 GHz quad-core)
RAM: 12GB DDR3
Video: NVIDIA GeForce GTX 970 w/ 4GB RAM (PCI Express)
Linux Kernel: 4.8.0.54-generic
NVIDIA video driver: 375.66
X-Server: 11.0
Renderer: GeForce GTX 970/PCIe/SSE2
OpenGL version: 4.5.0 NVIDIA 375.66

/etc/profile:
  export __GL_YIELD="USLEEP"

Kwin compositor settings:
  Rendering backend: OpenGL 2.0
  Tearing prevention ("vsync"): Automatic


** Other notes **
This issue was also present in plasma 5.9 and possibly earlier, but I didn't really notice it until I was using plasma 5.9.

Bug #374555 appears to describe the same issue, but was closed due to lack of information. Please let me know what other info/testing you would like me to provide.