[eluser]WanWizard[/eluser]
Depends on the platform, whether or not the webserver and the database run on the same platform, the number of page requests (concurrent connections), etc. So, very difficult to say.
I'd say it's going to be noticable when you have a web farm talking to a database farm, and you have a lot of concurrent connections. Just ran a test with ApacheBench on my dev server (1000 requests, 20 concurrent), and I can hardly see a difference in the average statistics.
What you could try (don't know if it will work), is use a persistent connection for your primary database, and no persistent connection for the second one.