Its prime time

[mdettweiler]

Originally Posted by Mini-Geek

It will probably only take about a day. My most recent Top 5000 prime (http://primes.utm.edu/primes/page.php?id=84463) verified in 304 seconds, I discovered it in 349 on my 2.5 GHz Athlon. A timing of this large prime comes to 1.17 days. Taking the same proportion, his should take about 1.02 days, or about 27 minutes over a day, or about 11:03 tomorrow morning. I think this is in Prof. Caldwell's local time? Which is CDT (GMT-5), right?
I'd say it's not too long.

Congrats on the megaprime!

One difference, though: Because PSP's latest megaprime is a Proth prime, rather than a Riesel one like yours is, the Prime Pages will verify it with PFGW, rather than with LLR. I don't know why they do that, but for some reason that's what they do--so, thus, it may take more than twice the time to officially verify this prime than it did to originally find it.

Maybe someone should talk to Prof. Caldwell about setting up his automated verification system to use LLR to verify Proth primes as well as Riesels?

[Sloth]

Originally Posted by axn1

BTW, Was this found on PrimeGrid or LLRNet?

Originally found using LLRNet. They verified with proth.exe

[mdettweiler]

Originally Posted by Sloth

Originally found using LLRNet. They verified with proth.exe

Proth.exe? Why not use LLR--it's much faster, and for base 2 numbers like the ones PSP is testing, it gives you a definite primality test for both +1 and -1 numbers, so there's no need to use Proth.exe.

[axn]

Originally Posted by Mini-Geek

It will probably only take about a day. My most recent Top 5000 prime (http://primes.utm.edu/primes/page.php?id=84463) verified in 304 seconds, I discovered it in 349 on my 2.5 GHz Athlon. A timing of this large prime comes to 1.17 days. Taking the same proportion, his should take about 1.02 days, or about 27 minutes over a day, or about 11:03 tomorrow morning. I think this is in Prof. Caldwell's local time? Which is CDT (GMT-5), right?
I'd say it's not too long.

What Anon said. By my estimate, it should take 3 days. Anyway, this is the kind of scenario for which there is the "External" classification -- big prime, found and carefully verified by projects.

[Mini-Geek]

I contacted Chris Caldwell regarding it.

> 'There'''s some discussion at mersenneforum.org
> (http://www.mersenneforum.org/showthread.php?t=10189) about
> the recently discovered megaprime'''s double check
> (http://primes.utm.edu/primes/page.php?id=84603,
> 265711*2^4858008+1). axn1 thought it would take ages, but I
> figured it would only be about a day. Would it be possible
> to give an ETA for the double check, and possibly accept a
> third-party double check if it'''s too long?'

My server will probably take just under a week. Who is offer a double check?

CC

It has passed a full, verifiable double check, no? If so, and whoever ran the double check can give him the details of it, he seems open to accepting it.

[ltd]

Before anouncing the prime here and on primepages the following tests were copmpleted:

The original test by Sloth using llrnet
Using llr test by Joe_o and VJS
Using PRP another test done by me. (This one to have a fast test using a different calculation)

Also a test using Proth was started on one of my machines to make sure that a completely different math package is used. Due to some real life issues that test is not finished yet. ( The machine was needed for some "real" stuff)

And to answer the question: Yes it makes sense to use different programms even slow ones as Proth.exe as they use diferent calculation methods. This makes sure that the number is really prime and not an artifact of a strange error in a programm.

As PSP is such a small project with a very low number of primes it never came into my mind to ask Prof. Caldwell if he would accept our tests as external prove. But with our primes getting very large now it could make sense to agree with him on a set of tests that satisfy his needs and take the load from his machine.

[axn]

Originally Posted by ltd

Before anouncing the prime here and on primepages the following tests were copmpleted:

The original test by Sloth using llrnet
Using llr test by Joe_o and VJS
Using PRP another test done by me. (This one to have a fast test using a different calculation)

These tests aren't as independent as one would think. In fact, they are all practically the same test (basically computing 3^(n-1)/2 mod n) using essentially the same FFT library.

PFGW perorms a N-1 test which uses a different algorithm (but same FFT library), where as Proth.exe performs the same algorithm (but different FFT library). In that sense, either of PFGW or Proth can be considered a truly independent test.

It is just a question of what is acceptable to Prime pages.

[m_f_h]

First of all: congrats to the lucky finder and the whole team!

Originally Posted by axn1

These tests aren't as independent as one would think. (...) It is just a question of what is acceptable to Prime pages.

Watch out, or some guys on this forum will detect another conspiracy about that prime...

at least I can say that its remainder upon division by 5 is 2, so it is not divisible by 5 :-)
In fact, one can even say that it ends in the digits (I like the ending digits...)
8124165274535163152784174078517754808463610920703724549019347635457900199302024330637815279586902017

and that it starts with the digits
35611077935304774640479579162325567520...

[VJS]

I think it's a big conspiracy that those numbers are not divisable by any other number but themselves and one.

Dunno, maybe they are just numerically greedy or something...

[Sloth]

I'll have to print it out to check. So how many characters per line and lines per page can fit so that it is easily read. And does anyone have a few reams of paper I can borrow??

[Joe O]

Originally Posted by Sloth

I'll have to print it out to check. So how many characters per line and lines per page can fit so that it is easily read. And does anyone have a few reams of paper I can borrow??

It can be done on one 29" x 40" piece of paper.
Contact this company if you are interested:
http://www.perfsci.com/souvenirs.htm#primes

The digits are very clear, but unless you have very, very good eyesight, you will need one of our watchmaker's loupes or an equivalent magnifier to read the digits.

And no, I don't know what it will cost.