I just noticed it, they do both the same… thats what i guess…
But my intuition says that there might be more behind it.
So what is the difference?
No collision makes the engine use only visual and spacial processing. No objects collide with it.
This can be used for useless backgrounds in games.
Ghost allows for physics (gravity only since it can’t collide with anything).
NinjAcademy
A Ghost “knows” when something is colliding with it, so if you hooked up a Collision sensor, it would register whenever something passed through it. Any type of object—static, dynamic, rigid body, etc.—can be a ghost.
No Collision, by contrast, means no physics calculations whatsoever. The object is static, and no collisions will register. You should use this option to save CPU time for objects that don’t need physics. Probably more commonly used.
Hope this clears it up.
EDIT: Simul-post, sorry. :o
Ghosts can be detected with Physics sensors (e.g. Ray)
Dynamic objects will travel straight through Ghosts rather than bounce back.
No-Collision can’t be detected by Physics sensors. It is the replacement of the “no collision” attribute at single faces from pre-2.5.
2 other were faster 
your Fast guys, and i Thank you very much for the clreaing up!