Well , simply put , the actual compute power of the GPU in this scenario is what will limit performance and the game will often hit these restraints before tapping into the entire memory buffer on either gpu. Not to say that Memory speed and bandwidth have no effect becuase they certainly do. The caveat with that is memory performance will stretch its legs on higher resolutions , the 850M while only 32GB/s memry bandwidth , to the 80GB/s (or so) will likey not be the hardware limitation with many , many titles. I would think the higher shader(ALU,CUDA Cores,SP..) count and texture unit count will get those pixels moving just as well or better than the desktop GTX 650