Benjamin Eberlei has posted some benchmarks he's put together as a part of his attempts to optimize his Zend Framework application.

I haven't changed great parts of the code or anything, I just benchmarked how different global include strategies affect the overall performance of my blog software written with help of the ZF. The following include strategies were tested (I've used the Zend Framework 1.6 RC2 package for all of them): Zend_Loader_Autoload (default and swapped include paths), ZF 1.6 RC2 stripped of require_onces and stripped of require_onces/no autoload and all files directly included.

He ran the tests both with and without APC enabled with a larger difference between the stats without it turned on for each of the four tests.