(M48) NEW MERSENNE PRIME! LARGEST PRIME NUMBER DISCOVERED!

[J.F.]

Originally Posted by Prime95

and now the bad news.....

Press coverage is always hit or miss. One of our least covered discoveries occurred when we sent the press release out on a Friday. Consequently, Scott is emailing our press contact list Tuesday at 6AM PST. We'll try to coordinate mersenne.org's web site, this forum, Chris Caldwell's site, and maybe wikipedia.

Ahh, that's surprising and rather disappointing... well you seem to have a good reason to do so.

[Ralf Recker]

The Wikipedia editing has already begun:

To help visualize the size of the 48th known Mersenne prime, it would  require 3,461 pages to display the number in base 10 with 75 digits per  line and 50 lines per page.

The number of digits is from the previous record holding prime:

3460*75*50=12975000
3461*75*50=12978750

and a new entry has been added to the table of known mersenne primes.

[ixfd64]

On the bright side, we have four more days for sleuthing.

[ewmayer]

Originally Posted by axn

After putting together all the clues in this thread, I've narrowed it down to

curtisc                    58******  C        Jan 25 2013
curtisc                    59******  C        Jan 25 2013

I'm guessing it is the first one.

My note said "without any knowledge of Primenet server traffic". Y'all are so busy parsing server logs, you've apparently forgotten how to do math. My point being, by trying the math-based approach, people might actually learn something useful about GIMPS-style computations.

[AG5BPilot]

Originally Posted by Batalov

There's also another cloud on the horizon. PGrid got more than probabilisitcally expected lucky with some sizeable GFNs. The wrGFN (world-record generalized Fermats) could get lucky, too. That's b^4194304+1. Their chances are as good as ours, aren't they?

At PG, we're searching around 16.5 million digits right now, and the tests take about 5 days on a GTX 580.

That's pretty good, but we've run into reliability problems that seem to be related to the size of the numbers and/or new video cards and/or new video drivers. That's hurting the overall GFN search on the N=4194304 numbers. The program won't start on a lot of individual GPUs. Needless to say, that hurts the effort significantly.

If some of the guesses I've seen here are accurate, at the rate we're progressing it may be late 2013 before we're searching above your new record. So you can relax and take the rest of 2013 off. You've earned it! :)

Mike

P.S. Repeating what I said on our forums: congratulations on the fantastic find! Well done and kudos to all involved.

[pinhodecarlos]

Originally Posted by AG5BPilot

At PG, we're searching around 16.5 million digits right now, and the tests take about 5 days on a GTX 580.

That's pretty good, but we've run into reliability problems that seem to be related to the size of the numbers and/or new video cards and/or new video drivers. That's hurting the overall GFN search on the N=4194304 numbers. The program won't start on a lot of individual GPUs. Needless to say, that hurts the effort significantly.

If some of the guesses I've seen here are accurate, at the rate we're progressing it may be late 2013 before we're searching above your new record. So you can relax and take the rest of 2013 off. You've earned it! :)

Mike

P.S. Repeating what I said on our forums: congratulations on the fantastic find! Well done and kudos to all involved.

I think you have classic heat transfer issue on the GPU's running the search but that's my humble opinion.

[AG5BPilot]

Originally Posted by pinhodecarlos

I think you have classic heat transfer issue on the GPU's running the search but that's my humble opinion.

I agree, but that's a different problem entirely, and one we understand fairly well. The problem I was talking about is an initialization problem that occurs before the crunching even begins.

Anyway, I don't want to detract from the joy here by talking about my CUDA problems. Feel free to pop over to PrimeGrid to continue, if you want.

[NBtarheel_33]

Originally Posted by Prime95

and now the bad news.....

Press coverage is always hit or miss. One of our least covered discoveries occurred when we sent the press release out on a Friday. Consequently, Scott is emailing our press contact list Tuesday at 6AM PST. We'll try to coordinate mersenne.org's web site, this forum, Chris Caldwell's site, and maybe wikipedia.

What about contacting CNN (they call it an iReport or some such)? Or our local news stations/newspapers? And don't forget your wager with the fellow @ NPR!!

[ixfd64]

Originally Posted by rcv

Ernst said Serge is using an FFT size of 3328K to feed 26 cores. This suggests the number is above 57.2M.

Not necessarily. It's possible to test a number using a FFT that is far larger than one designed for said number.

[Uncwilly]

Originally Posted by ixfd64

Not necessarily. It's possible to test a number using a FFT that is far larger than one designed for said number.

Yes, especially when you are near the crossover it is safer to do so.

[philmoore]

Originally Posted by Prime95

and now the bad news.....

Press coverage is always hit or miss. One of our least covered discoveries occurred when we sent the press release out on a Friday. Consequently, Scott is emailing our press contact list Tuesday at 6AM PST. We'll try to coordinate mersenne.org's web site, this forum, Chris Caldwell's site, and maybe wikipedia.

George, I think you missed a good bet by not coordinating with Groundhog's Day: New Mersenne prime sees its shadow, goes back to its hiding place for another 4-year drought, or else it doesn't, in case we get another new discovery in the next 6 weeks!

So what is the weather prediction for tomorrow?