Thanks for the reply.
If I the way I'm understanding is correct then it is pretty slow compared to a bounding box method..
The way I see it is that it uses a dot product per node in the tree except the leafs, to check if the camera is in front or behind the nodes..
But if what I'm saying is true it wouldn't explain the popularity of the method, there must be some other trick I'm not seeing here..
It would still be nice to get some more replies if that's possible?