The GMP program for 2^107-1 recorded a small negative time... but in fact took about 45 32-bit Athlon GHz-hours:

[CODE]Exponent Total number of zeroes
2 1
3 1
5 2
7 5
13 8
17 13
19 15
31 28
61 96
89 140
107 185
[/CODE]

I am not going to attempt Lucas' 2^127-1 :wink:

[QUOTE=paulunderwood]The GMP program for 2^107-1 recorded a small negative time... but in fact took about 45 32-bit Athlon GHz-hours:[/QUOTE]
Oh, yes there is a problem in timing routine above about 1800 sec ( because I've used double variable ). But the program is good.

[QUOTE=casevh]
89 * ln(2) * [ln(ln(sqrt(2^89-1))) - ln(ln(3))]
[/QUOTE]

I don't know what's going on. I evaluated that on Mathematica and I got 206, not 326...