In my opinion, those are some very unreliable benchmarks. Firstly, you should know that they probably included the whole ZFW package in a script and benchmarked it, this is to make a general estimate of the framework's performance and should not be trusted for production use. Did you ever see a professional include the whole ZFW in their bootstrap? Well, I have neither.
Secondly, I don't see in any way that the performance could drop when using APC caching. APC is supposed to cache pages/content into shared memory and prevent them from being reparsed every request. I have no idea how they performed the benchmark but this sounds very impossible. And the benchmark package just contains a simple controller so it doesn't really show how they did it. You should do some benchmarks of your website(s) without and with APC caching and see for yourself. Of course, only use APC where it can be of use. I bet they probably benchmarked with "keepalive off" on purpose or ran several system virus scans while benchmarking the ZFW :-P -- Isaak Malik Web Developer
