It would use CPU instead of GPU as it would not find the nVidia card to process with. Frames per second capture and live capturing would be to where you would have to scan slower in order to capture everything without losing tracking.
Rule of thumb is 15 fps is very slow to maneuver with by going slow, 30fps is easy to not lose tracking..
if high or very high (live tracking) craps out, try going to medium, and scanning to get more fps
then you can later go to the reconstruction tab and reconstruct a higher point cloud/model from the data
The more fps you have, the better you can go back later and tweak the same scan if you need.
the quality is all about how you scan, and which you use for live/reconstruction for the point cloud creation.