If you want to move a mountain, you'll need to...

[Oddball]

...move it one rock at a time.

In an attempt to revive this project, I'm proposing a new goal: test n=390,000 from k=1-150G. n=390000 is twice the size of the project's original n=195000, and that range will give us a 90-95% chance of finding a twin.

I know, some of you may be skeptical because I'm new here, and the fact that I'm suggesting n=390000 instead of the previous sieve effort (n=480000-500000) probably won't help things. But please be open to my suggestion. It doesn't make much sense to test n=480000-500000 from k=1-10M because the time savings for testing smaller k's are already gone once you get above k=100,000. A test of, say, 123456*2^500000-1 will take the same time as a test of 9,876,543*2^500000-1. To make things worse, the largest time savings occur for low k's (k=1-2000), and they have all been tested by RPS and NPLB for the n-range already sieved.

In the end, testing k=0M-10M will be less than 0.5% faster than testing k=100M-110M if the sieve depths are equal. But if we spend our sieving resources on one n instead of on 20000 n, we can go to a much higher depth that is more than enough to cancel out the small 0.5% benefit from testing low k values.

Finally, there's no need to rush to really high n levels now. Finding a twin for the smaller n=390000 is expected to take almost twice the amount of time it took to find the n=333333 twin. It may take even longer because Primegrid may not want to help us this time, and if Primegrid does want to help, the boost may not be as much as last time because their efforts are spread out across a whole bunch of projects now. Also, the 5000th largest prime is 150,540 digits long, so finding a non-twin prime for n=480000-500000 won't get you on the top 5000 list anyway. And after a twin for n=390000 is found, we can always test n=480000-500000 if that's what most people on here want.

It's really a shame to let this project die out. Most projects like GIMPS, 17 or bust, and the Woodall Search have a long completion time per candidate, and your chances of finding a prime are extremely small due to the high n-values. Projects like RPS and NPLB don't have this problem, but they don't offer the glory of setting a world record either. Three years ago, this project offered the best of both worlds - short testing times, a good chance to get on the top 5000 list, and a dream of a world record. Getting on the top 5000 list may not be possible now, but the appeal of short testing times and a potential world record remain. Why not take a chance and join us?

[Oddball]

Like the title of this thread says, we can tackle a big problem by breaking it down into small, manageable parts. The range reservation thread is located here:

http://www.mersenneforum.org/showthread.php?t=13376

I've done the first part: k=1-300,000 has been tested with no primes (single or twin) found.

[Historian]

Originally Posted by Oddball

Three years ago, this project offered the best of both worlds - short testing times, a good chance to get on the top 5000 list, and a dream of a world record. Getting on the top 5000 list may not be possible now, but the appeal of short testing times and a potential world record remain.

You didn't mention it, but at one time, this project also got onto the top 20 list for arithmetic progressions of primes.

[agent1]

Originally Posted by Oddball

Three years ago, this project offered the best of both worlds - short testing times, a good chance to get on the top 5000 list, and a dream of a world record.

You didn't mention it, but at one time, this project also got onto the top 20 list for arithmetic progressions of primes.

tps also found biggest known chen prime:

http://en.wikipedia.org/wiki/Chen_prime

Getting on the top 5000 list may not be possible now

operation megabit twin?