[quote=Mini-Geek;191915][URL]http://www.mersenneforum.org/showthread.php?t=9675[/URL][/quote]
thanks
has anyone worked out a formula for the chance of finding n primes?

448-464 is complete.

Congratulations to the project, and also to Flatlander! I find this project interesting, not only because it is a generalized Sierpinski/Riesel problem, but because it also appears to be just a little bit more difficult than the problem we are pursuing at "Five or Bust". That project was launched just after I discovered a 358,640 digit probable prime number, so we started our project with five unknown sequences at about the same point where you have six. At this point, "Seventeen or Bust" still had twelve sequences left, so we can hope that both of these problems are ultimately easier. Only more testing can show for sure!

[quote=henryzz;191920]thanks
has anyone worked out a formula for the chance of finding n primes?[/quote]

I have not. I could probably figure it out for 2 or 3 but to make it variable as in n primes like you're suggesting would require a lot of thought to get it correct in all circumstances.

[quote=philmoore;191988]Congratulations to the project, and also to Flatlander! I find this project interesting, not only because it is a generalized Sierpinski/Riesel problem, but because it also appears to be just a little bit more difficult than the problem we are pursuing at "Five or Bust". That project was launched just after I discovered a 358,640 digit probable prime number, so we started our project with five unknown sequences at about the same point where you have six. At this point, "Seventeen or Bust" still had twelve sequences left, so we can hope that both of these problems are ultimately easier. Only more testing can show for sure![/quote]

Thanks Phil. I too have made a reference or two in another thread comparing the difficulty of proving your project to proving this particular CRUS base. Yours "should" be far FASTER to prove for 2 reasons: Base 2 tests much faster for the same decimal size and base 6 is simply a higher base, which gives less possible candidates within each respective decimal size range. That said, the n-value at which each project finds its final prime may be somewhat close, relatively speaking. I'm currently speculating a 50-50 chance for n=30M-60M on this one (only a SWAG at this point) and as I recall you mentioned a 50-50 chance for something like n=20M for yours.


Gary

That is a bit optimistic, as my estimate was that Five or Bust had a 20.5% chance of finishing for n < 50 million. We have been lucky in finding three primes so far, but I don't see any reason that the luck can't continue!

What was the original n-level where you had a 50-50 chance when you started your project with 5 k's remaining? If it was 20% by n=20M then it was probably 50% for more like n=50M-75M.

If that is the case, then a "little bit" better estimate for a 50-50 chance of proving ours is more likely to be n=200M-300M.

In the near future, I think I'll take a closer look at the weights and chances of prime at various levels for each of our remaining 6 k's to attempt to narrow it down a little better.

The file for n=500K-750K sieved to P=43T has been posted! There's a ton of work in there so come and get it. :smile: Let's see if we can bring it down to 4 or 5 k's remaining (or less!) at n=1M.

Keep in mind that the file now contains all 6 k's remaining so will take quite a bit longer than the n=370K-500K file that only had 4 k's initially and subsequently 3 k's remaining.

Note for everyone: When reserving a range, please make sure that you take from xxx001 thru xxx000 (as opposed to xxx000 thru xxx999). For example, for an n=505K-510K reservation, you'd take n=505001 thru 510000. It's a small point and a majority of ranges won't have the very specific starting and ending exponent that would make a difference but I'd hate to miss a single test and a huge prime.

Everyone can give your input on beginning to load part or all of the ranges in a PRPnet server. With tests this long, I'm fairly confident in the stability of one. We could do it "NPLB style" and load n=~20K ranges in a server while leaving ranges not in the server open for manual reservation. On the other hand, we can just continue with all manual reservations for a while. That has worked quite well so far.


Gary

I'll continue with 500-510K.

Taking 510-514.

57023*6^483xxx-1 is 3-PRP! (3729.6409s+0.0354s)
:w00t:
...veryfying now!