Being able to run two images per second through a post-processing and/or computer vision pipeline still sounds pretty useful to me.

Shoot, Blender still supports users without accelerated graphics. (Or did five years ago, things may have changed.)