VirtualDub capture: Overlay vs. Preview

[^rooker]

When capturing in VirtualDub, there are 2 different methods to choose for handling the video displayed during capturing:

1) Video > Overlay (O)
2) Video > Preview (P)

I always assumed that "overlay" would use hardware acceleration of the graphics card by using a regular video overlay to display the incoming video directly and therefore cause less CPU load.

Wrong!

I've now had the problem with visual statics in the capture view, which caused me to experiment with the view settings.

Switching to "Video > Preview" did not only completely remove the visual static noise (white drop-out like artefacts) from the live view of the capture mode, but also had some interesting advantages, when choosing "Video > Preview acceleration":

1) You can choose between different interlaced/progressive, field, etc. views - in realtime!
2) Choosing any other option than "Off" reduces the CPU load

Exactly the opposite way I expected! So on my setup (Dual-core i5 3,3GHz), it's as follows:
*) Using "Video > Overlay" had ~10-12% CPU load while viewing (no capture)
*) Using "Video > Preview" / "Preview acceleration > Interlaced frames": ~1-2% CPU load