I think it is easier on the MySQL database to use the COUNT * ...-idiom to test whether a record exists, rather than relying on a failing insert because the primary key already exists.

The docs of MySQL state that COUNT * ... has been specifically optimised for such things.

Update:On second thought, if you have to insert the new data anyhow, then perhaps it is more resource-effective to rely on the "primary key already existing"-error, provided that most of the inserts are likely to succeed and such errors are rare.