In that particular case, ack's performance advantage seems to arise mostly because of a bug that causes GNU grep to be terribly slow in multibyte locales:
http://blog.i-no.de/archives/2008/05/07/index.html http://savannah.gnu.org/bugs/index.php?14472
It doesn't use "perl compatible regexes", it uses Perl.
I would argue that 'perl regexes' are a member of the set of 'perl compatible regexes' ;-)
But, it's faster even when doing the same thing. For example, see the benchmark at http://blog.i-no.de/archives/2008/05/06/index.html.
Then end result is that, on the same search, ack is approximately 5 times faster than grep. I ran a few test cases myself and got similar results.
But the real reason I like it is the perl compatible regexes. The differences between perl regexes and gnu regexes are just big enough to be an annoyance.