Benchmark it under real conditions. There are a lot of conditions that affect the speed of the databases. Just a silly example:

If I wanted to "prove" that Pg is slow as hell, I would run two tests: One on an in-memory table in SQLite on a fast machine with lots of RAM, and one on a permanent table in PgSQL over an ISDN connection to ancient hardware with as less RAM as possible and slow disks.

If I wanted to "prove" that SQLite is slow as hell, I would reverse the conditions: A permanent table in SQLite, stored on an uncached floppy disk drive vs. a temporary table in PgSQL on the fattest server I could get, connected with at least gigabit ethernet (or via local loopback), and with PgSQL tuned as much as possible.

Alexander