acutally ive read up a bit on this. converting to subdivided is using NURBS. (some1 correct me id im wrong on this,) basically subdividing the faces multipies the lines which gives a bigger file and more of a render time. using convert to subdivided uses nurbs wich i guess condenses your model and manipulates it.

NURBS, short for nonuniform rational B-spline, is a computer graphics technique for drawing curves. A NURBS curve is defined by a set of weighted control points, the curve's order and a knot vector. NURBS are generalizations of both B-splines and Bézier curves, with the primary difference being the weighting of the control points which makes them rational (non-rational B-splines are a special case of rational B-splines, in practice most NURBS curves are non-rational).

thats what i found online but if you look at your model in point edit mode with no shading (wireframe) you can really see what it does. once you know what it does you can kind of pick and choose if you want convert to subdivided or subdivide faces.

i usually subdivide faces when im at the very start of my model and only when i really need that many lines.

but i think if you have a finished model or close to it. you dont want doe divide the faces you would want to convert to subdivided faces.