In a, all 52 cards are sorted through once, to place into 4 piles by suit. This is 52 comparisons. These suit piles are then sorted. I will assume that these undefined sorts take place at n, as if a line of unsorted cards was on the table, and a new line of sorted cards were pulled out and placed above. This creates an additional 4*13, or 52, comparisons. Joining these 4 piles gives a sorted deck, at a cost of 104 comparisons.
In b, the cards are dealt into 13 piles initially, resulting in 52 comparisons. These 13 piles, stacked back together, and sorted again, result in another 52 comparisons to get into their final order.
In c, the cards would be sorted essentially as a selection or bubble sort, resulting in a possible 2704 comparisons, when compared relatively for proper position.
This makes method a and b the fastest. A better solution, however, may be to lay out a grid or 'hash' of all 52 possible values, and place each card into its sorted order. This would result in 52 comparisons for a sort.