Simulate Physics and Physics Constraint


In the following two case, physics constraint are applied on each components relatively to the root component (Tx, Ty, Tz, Rx, Ry, Rz locked) :

Does simulate physics on all components of an actor is less efficient than simulate physics only on the root component ?

What are the pros/cons with these cases ?

Because is the first case, i can get component relative location (by linetrace), but not in the second (root location instead).