Pear vs Phplib vs adodb

I'm looking to choose a set of library code principally for database functions. So far checked out Pear, Phplib and adodb which all appear to be well crafted and easy to use. Irrespective of differences in the actual functions provided (which of course are well documented) can anyone briefly compare and contrast them from experience working with 2 or more such libraries in a production environment in terms of robustness and speed (I appreciate that speed tests have put adodb top, then phplib and finally Pear, but have people found such differences to be very noticeable and if so under what specific circumstances?) Cheers, David Eisenhart