Well, poorly used crease creates unrealistic edges. So do poorly used edge loops. Edge loops allow more control of the edge, but at the cost of more control points, more verts, and one of the main reasons you’re using subsurf anyways is to cut your base vert count-- what’s always going to give you the most control, for a given vertex budget, is not using subdiv at all. Crease can deform differently than edge loops, but in my opinion, better: you have to weight your control loops carefully to prevent them from deforming poorly, and you’re talking about the difference between creating verts before armature vs. after armature, when I think most people would much prefer to create their verts (via subdiv) after. Creases can’t be exported, but realistically, neither can live subdiv (although it probably depends on destination, and recent changes to Blender have improved compatibility with other engine’s subdivision algorithms), so if you want to export, you’re writing your subdiv anyways, at which point your creases don’t matter, they’re already written into the mesh.
What I think is that doing subdiv modelling with the use of creases requires a good knowledge of topology, and most of the people who want to do hard surface modelling don’t want to worry too hard about topo-- they want to use more verts to make up for their topo problems. I’m talking things like non-planar 3/5 poles mostly, at extrusion points, but you also have the hard surface boolean crowd. If you want to use creases, your non-planar 3/5 poles will bite you in the ass. If you don’t care about vert count, and you’re not animating, use control loops instead, they’ll let you get away with more, without having to think as hard.
Not really a materials question; you might get better responses on the modelling forum.