It depends on what size of exponent you're working on; a 3.2GHz P4 manages an iteration time of 0.029 for numbers in the 18-million range, a 2.4GHz Athlon X2 would take 0.17 seconds for numbers in the 60-million range.

The prime95 code happens to run a good deal faster for exponents of the same size on Intel than on AMD chips of the same GHz.

http://www.mersenne.org/report_benchmarks/