Let's benchmark your Windows PC against my Mac. As the challenged party, you are customarily entitled to propose the test - but I would suggest the usual suite of computationally-intensive benchmarks. We'll run the tests, compare normalized results, and determine the performance ratio per CPU cycle.
I have a PowerMac with dual 2.0 Ghz IBM G5 processors running Mac OS X 10.3.7. I'm interested to see how it compares against your machine.
I'll be compiling a BSD kernel while you decide.
Won't tell us very much about the G4, will it?