Naive fibonacci in pypy

As this is my last post this year, I'll just keep it short. I've been busy doing all sorts of things, so here's just a fast revisit to our old friend, the naive fibonacci algorithm, this time featuring Pypy. I downloaded the 1.0.0 version and ran the by now famous naive fibonacci. To get to the point, it took 184 seconds. Yep, you read it right---3 minutes. The same thing that took 27 seconds in CPython 2.5, or around 3 seconds using Psyco. Now, I didn't read much about pypy, but as I understand it, it should be faster than the CPython implementation... And AFAIK, the guys that made psyco abandoned that project to go work on bringing similar functionality to pypy. Now, I hate to bitch (no, no I don't:) ), but this really seems kinda slow. What's going on here? It's probably the function call overhead that slows it down. One should look at the code of pypy and see what exactly they put on the stack each time... The point being, I'll stick to CPython for a while more, at least until the naive fib runs under 20 seconds :) ('Cos what's more important in life than running a fast fibonacci algorithm?)
Happy holidays everyone