Delayed 2D Erase brush response on Windows with hybrid Intel/NVIDIA graphics

Hello,

Over the past several days, Slicer has developed a noticeable delay when moving the Erase brush over a 2D image slice. The brush follows the mouse a fraction of a second late. Editing itself can work, but the movement is still laggy. Reinstalling Slicer did not resolve it.

Environment:

  • 3D Slicer 5.12.4, 64-bit application.
  • Windows 11 Pro 26H2, build 26300.10075.
  • Dell XPS 17 9720, 32 GB RAM.
  • Intel Iris Xe, driver 32.0.101.7088.
  • NVIDIA RTX 3060 Laptop GPU, now driver 610.74 / 32.0.16.1074.
  • External Dell S3222DGM monitor.

The NVIDIA driver was updated from 596.08 through Windows Update. I retested afterward and the 2D brush delay remains. There has also been occasional external-monitor blinking and typing-focus loss outside Slicer, so I am not assuming this is exclusively a Slicer bug.

Checks so far:

  • Before the NVIDIA update, a blank startup with temporary settings and the startup script ignored succeeded. Startup with saved settings also succeeded.
  • A simple synthetic VTK render reported the RTX 3060 and OpenGL 4.5 and completed successfully before the driver update. This was a rendering check, not a Paint/Erase performance test.
  • A separate episode where clicks had no effect involved Margin being active and the editable area being restricted. Selecting Erase restored editing, but did not resolve the cursor/brush delay.
  • No clean Sample Data reproduction, measured frame rate, or controlled single-slice versus multi-view comparison has been completed yet.

I saw the forum’s highlighted screen-glitch notice about NVIDIA 595.71. My current driver is 610.74, so I do not know whether that is related.

What is the most useful next diagnostic to distinguish brush/rendering overhead, scene-specific complexity, and a system/driver issue? In particular, should I compare a single slice view, Show 3D off, and Data Probe magnification/crosshair settings? Is there a recommended local timing/profile procedure or public sample dataset for this?

No patient data, screenshots, scene files, or raw logs are attached. I can provide focused, non-identifying technical details if needed.

Thank you.