I am trying to do the common centroid matching for transistors in the diff. pair (each transistor has W=12u, L=3u, Multipliers = 12). What is the better common centroid matching?
Both are good. I'd leave the decision to the aspect ratio which is best appropriate to fit into the cell design. The 8x3 array will achieve an aspect ratio closer to a square form (vertical gates), which rather supports best matching.