KDE Overview causing VRAM spikes and crashing KWin (GTX 1060 3GB, Wayland)

Help
,
1

Hey,

I’ve been experiencing occasional crashes when triggering the Overview effect on KDE Plasma (Wayland).

What happens:

  • Screen freezes briefly

  • Desktop session restarts (KWin crash)

I monitored VRAM using nvidia-smi:

  • Idle: ~600 MB

  • Triggering Overview: jumps to ~800 MB (and then crash)

But I believe the real issue is a very fast VRAM spike that doesn’t fully show up in nvidia-smi (since it updates every second).

Logs show:

  • GL_OUT_OF_MEMORY

  • kwin_wayland core dump

So it looks like KWin fails to allocate buffers during the Overview animation.

My guess:
Overview creates multiple window textures at once → VRAM spikes → allocation fails → compositor crashes.

Setup:

  • GTX 1060 3GB

  • NVIDIA 580.142 driver

  • Wayland session

  • CachyOS

Is this expected behavior on 3GB GPUs?

Would love to hear if others with low VRAM GPUs have the same issue.

2

Under most of what I constitute as “normal” use I consume 5-6gb of vram all the time watching nvtop all day with htop in a corner of my screens. Just kwin_wayland alone right now is using 1.3gb alone (and that is light for normal), firefox another 1gb, wave terminal 500mb, not even including my normal steam gaming, and this is after a fresh upgrade a few weeks ago. Usually after a month or two it’s far worse averaging ~8gb usually I’d say, as I can only usually load around 6-7gb llm models without popping something on my 16gb 3080.

I’m actually running in xorg/xwayland mode on a fluke after an arch upgrade currently, where after 2 weeks is running FAR more noticeably smoother and less memory hungry (both vram and ram) than wayland has after using it for the past few years now. I was going to switch back to wayland after the work day not even noticing except that videos played much crisper/higher fps in youtube suddenly, and so did games that finally I was like wait a minute… Let me check something… xorg? How did I get back on xorg? Guess sddm default back and didn’t notice.

It’s been refreshingly… better than wayland? So much that I stayed on xorg. (please don’t stone me for heresy)

Hate to say it, but that 3gb of vram has probably outlived its usefulness these days. You might have better luck with xorg honestly, or see what is using your vram with nvtop.

3

After the fifth attempt, KWin crashed.

KWin crashed after repeated use of the Overview effect on Wayland. The crash is preceded by NVIDIA Xid 32 errors in the kernel logs, which suggests it may be related to the NVIDIA driver.

NVRM: Xid (PCI:0000:2b:00): 32, pid=1425, name=kwin_wayland, channel 0x00000088 intr 00040000
Process 1425 (kwin_wayland) dumped core.
#0  EglContext::shaderManager
#1  EffectsHandler::renderOffscreenQuickView
#2  EffectsHandler::paintScreen
4

I’m using arch not cachy, and not seen this sort of behavior as a close-enough-ish thing. I’m using currently nvidia-open 590.48.01-17 current with 6.19.9-arch1-1 kernel, plasma 6.6.3, and no issues like it with my 3080.

You’re sure it’s loading proper nvidia drivers and such? Like doing vkinfo, should see something like this:
$ vulkaninfo
Device Properties and Extensions:
=================================
GPU0:
VkPhysicalDeviceProperties:
---------------------------
apiVersion = 1.4.325 (4211013)
driverVersion = 590.48.1.0 (2475425856)
vendorID = 0x10de
deviceID = 0x249c
deviceType = PHYSICAL_DEVICE_TYPE_DISCRETE_GPU
deviceName = NVIDIA GeForce RTX 3080 Laptop GPU

Or nvidia-smi:
$ nvidia-smi
| NVIDIA-SMI 590.48.01 Driver Version: 590.48.01 CUDA Version: 13.1 |

I’ve seen stranger things than finding nouveau drivers running after an upgrade.

What does OBS say about your gpu too?

5 
❯ nvidia-smi
Sun Apr  5 04:34:50 2026       
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 580.142                Driver Version: 580.142        CUDA Version: 13.0     |
+-----------------------------------------+------------------------+----------------------+