"square is the best" - as pavanks told above - in terms of least mismatch. But there might be a higher priority: the block should fit into the full chip as to achieve a minimum die size. So an optimized global floorplan of the chip could advice an aspect ratio different from the 1:1 ratio.
If you have a standard cell type arrangement there is
often more value in laying out to a single-rack height
(conserving routing channels) than in making it look
square by occupying multiple racks (and rack pitch
may change later driven by higher level routing).
On high pin count designs you may not care at all
about core packing density (I/O-bound) or the aspect
ratio of sub-blocks.