On the theory that at least one other person is interested in the computational effort required to go after the EFF prizes for the larger prime numbers, I made some rough extrapolations from George's "Status" table and came to the following conclusions (generally to one significant figure): Ten-million-digit prime On the order of M33000000, FFT size 1408, 7 P-90 seconds per iteration, 8 P-90 years per exponent. One-hundred-million digit prime On the order of M330000000, FFT size 3072, 80 P-90 seconds per iteration, and 800 P-90 years per exponent. If I postulate the existance of a 900Mhz Pentium, the computation time for a LL test drops to merely 80 years per exponent. And, finally, the giga-digit prime: M3300000000, FFT size 4736, this time postulating a 90Ghz Pentium (three orders of magnitude faster than the current reference), the single iteration time is a "mere" 0.8 seconds per iteration, but since the algorithm requires ~3 billion passes, the per-exponent test time is still 80 years. These times are all PER EXPONENT; historical evidence suggests many exponents would have to be tested for each prime found. I would imagine the EFF folks could probably put a dollar in a savings account now, and it will grow via interest to fund the largest prize by the time it's claimed. ;-) Regards, John Gilmore (Not the one at EFF) ________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
