positive LL test?

[ewmayer]

Originally Posted by Prime95

GTX 570 concurs -- not prime. Well, at least it provided a few days of excitement here.

And you got a verified result in the end, so it's not like the runtime was wasted.

For the curious, a link to the exponent in question:
http://www.mersenne.org/report_expon...5520823&full=1

p-1 = 2.3.17.642361, so based on the famous extended crank p-1 smoothness hypothesis (eCPM1SH^TM) I would have judged the odds of that exponent yielding an M-prime as low - using the statistical measures of p-1 smoothness vs nearby primes I defined in post #32 of this related thread

B -smoothness: 65520823 is 821 of 1103, percentile = 25.66
S_{L[sup]2[/sup]}-smoothness: 65520823 is 820 of 1103, percentile = 25.75

(Compare these %iles with the ones in the summary table in post #59 for known M-primes with p > 100000).

[cardogab7341]

Originally Posted by ewmayer

It would be interesting to chart realized flops/watt for a mature program like cudalucas over time, to see if the ratio of touted/realized is at least narrowing from one GPU generation to the next, or if folks are simply throwing more silicon (i.e. watts) at the problem.

http://www.mersenne.ca/cudalucas.php

This is what you seek....

[ewmayer]

Originally Posted by cardogab7341

http://www.mersenne.ca/cudalucas.php

This is what you seek....

Thanks - so the dflops/W trend is actually quite favorable.