I remember seeing this, and ending up switching back to count(*), but I cannot remember why - it may have been for accuracy problems, and I am not sure which version of SQL I was using at the time (I know it was 2000 or 2005). I also wonder if Microsoft would have the query engine do a table/index scan if it doesn't really have to - select count(*) FROM '1 table' is a relatively common operation. My gut tells me if it is doing a scan, that is because it actually *needs to* do a scan. Of course if you don't care if you occasionally see 4,815,162,341 rows when there are actually 4,815,162,342 rows, then this is a good trick to know.