i started using count(1) when count(*) appeared not to return results in a timely manner (on very large table, lots of rows/columns). if truly implemented the same way by sql server, it must be serendipity i experienced a difference in performance.
that could be due to trying first with count(*), getting the table loaded in memory in the process, then moving on to using count(1). i think of '1' taking up less memory internally to process, but there's no real foundation for that.