count * performance issue

count(*) tooks much time...

but with the where clause we can make this to use indexing,... what where
clause we can use??

Am using postgres 7.4 in Debian OS with 1 GB RAM,

am having a table with nearly 50 lakh records,

it has more than 15 columns, i want to count how many records are there, it
is taking nearly 17 seconds to do that...

i know that to get a approximate count we can use
SELECT reltuples FROM pg_class where relname = TABLENAME;

but this give approximate count, and i require exact count...