When you need a database, you typically chose one according to its performance, and/or feature set. However, a "highly scientific performance test" might not speak much about your application's end resulting performance. Due to these reasons, I chose to create a "fully fledged app" using Magic and Hyperlambda, and I chose to create it twice. Once using Oracle's Sakila database and then once more using PostgreSQL's Pagila database.
The reasons why this is interesting, is because the Pagila database is more or less an exact "port" of the Sakila database, including each row in its tables. This allows us to measure the performance differences between these two different databases, and end up with a result resembling the performance difference we might expect in our application.