Holy new Mersenne prime, Batman! (M47 related)

[retina]

Originally Posted by sylvester

The way the currencies and central banks are doing nowadays would probably mean that in 15 years $150K will be just enough for a couple of postcard stamps and maybe a pitcher of beer.

Unless you live in Zimbabwe, where $100 billion would buy you 3 eggs.

In the just the last 10 days the currency has fallen to 1/10th. If that continues that would seem to make inflation 10³⁴% p.a.

[T.Rex]

Aug23th : 93.06% done. Never so close !
Sept6th : 67.50% done. Soon.

OK, OK. I know you don't care... You prefer to discuss with Bob...

Tony

[davieddy]

Originally Posted by T.Rex

Aug23th : 93.06% done. Never so close !
Sept6th : 67.50% done. Soon.

OK, OK. I know you don't care... You prefer to discuss with Bob...

Tony

We DO care Tony and how is Jeff getting on?

[T.Rex]

Originally Posted by davieddy

and how is Jeff getting on?

No news. But I'm home now, writing a paper for a French Scientific newspaper. 4 to 6 pages about Mersennes, LLT, GIMPS, M45 and M46 . Waow !
Sure that Bob would disagree that I talk about programs and machines I haven't designed nor built, but I spent so many hundreds of hours on this subject that I guess I can say interesting things...
About the 100M prime number, my guess is that it will be discovered in 2023 (if there is still petrole...).
Tony

[uigrad]

Originally Posted by T.Rex

About the 100M prime number, my guess is that it will be discovered in 2023 (if there is still petrole...).
Tony

At the rate of one prime a week, it should come much sooner than that. Speaking of which, shouldn't M47 be due any day now? ;)

Regarding oil reserves, if you only count the 12 biggest reserves, then there is enough oil available now for 65 years, and these reserves have only risen over the last 10 years. These numbers do not include oil shale or some offshore oil, which may be the largest sources available.

http://en.wikipedia.org/wiki/Oil_res...erves_in_order

If hardware continues to follow Moore's law, in 2023, the average pc will be 1000 times faster than today. I'd be rather surprised if we were still looking for a 100-digit prime at that time.

I'm actually more curious about when we will be done double-checking every exponent that produces less than 10 million digits. My guess is that we have at least another 5-7 years before we know "the smallest 10 million+ digit prime".

[ewmayer]

Moderator Note: I am moving all the Silverman-vs-the-World sideshow/flamewar posts to their own special thread. I ask all the hijackers of the current thread to take their ideological soapboxes and megaphones over there.

Thank you for your cooperation.

[Mini-Geek]

Originally Posted by uigrad

At the rate of one prime a week, it should come much sooner than that. Speaking of which, shouldn't M47 be due any day now? ;)

If we kept getting one a week, that wouldn't make our leading edge push to 100M digits as quickly as you imply. It wouldn't directly change our leading edge a single bit, it would just mean that the 100M digit prime would be about M1000 (as in the 1000th prime, not as in 2^1000-1).

Originally Posted by uigrad

If hardware continues to follow Moore's law, in 2023, the average pc will be 1000 times faster than today. I'd be rather surprised if we were still looking for a 100-digit prime at that time.

Um...we're not still looking for a 100-digit prime, we're looking for a 100M-digit prime. And, assuming you really meant 100M-digit prime, actually, a PC 1000 times faster than mine would still take 7 days. When you consider Moore's "Law" isn't really followed, and that there's problems we'll run into pretty soon like the speed of light, not to mention how sparse the primes will be at that size, I think Tony's (T.Rex) estimates for a 100M digit prime is quite reasonable, if not optimistic.

[ixfd64]

Originally Posted by ewmayer

Moderator Note: I am moving all the Silverman-vs-the-World sideshow/flamwar posts to their own special thread. I ask all the hijackers to take their sh*t over there.

Thank you for your cooperation.

Thank you so much. I was hoping a moderator would split this thread.

[uigrad]

Originally Posted by Mini-Geek

If we kept getting one a week, that wouldn't make our leading edge push to 100M digits as quickly as you imply. It wouldn't directly change our leading edge a single bit, it would just mean that the 100M digit prime would be about M1000 (as in the 1000th prime, not as in 2^1000-1).

That all depends on which assumptions you take. The density of mersenne primes is an assumption based only on imperical results, but I'll keep it. The factors that I was expecting to float were a) # of participants to GIMPs, and b) the luck of prime hunters in GIMPs.

You may think that I should have been more specific. If I was trying to make a point, I should have been more specific, but I was merely making a joke. Expecting a new prime each week is so silly that I didn't feel I needed a winky, but I supplied one anyway (check my original post).

Originally Posted by Mini-Geek

Um...we're not still looking for a 100-digit prime, we're looking for a 100M-digit prime. And, assuming you really meant 100M-digit prime, actually, a PC 1000 times faster than mine would still take 7 days. When you consider Moore's "Law" isn't really followed, and that there's problems we'll run into pretty soon like the speed of light, not to mention how sparse the primes will be at that size, I think Tony's (T.Rex) estimates for a 100M digit prime is quite reasonable, if not optimistic.

Mea culpa. I did in fact mean 100-million digit.

When I joined gimps in '99, people were saying Moore's law no longer applied -- we were too near the theoretical limits. Checking a exponent of 43M would have taken 1539 days with my P2-300Mhz, so Moore would predict that today it could be done in 15.39 days. I can't LL-test an exponent of 43M in 15.9 days with my Q6600, but I can LL-test 4 exponents of that range in 32 days, so the law has held up.

The theoretical limit of light will not apply if we develop molecular gates, or if we ever reach broad parallelization (ie neuro nets). I've seen "experts" predicting that Moore's law will hold until about 2020, but they're guessing just as much as any of us are. Since it is all speculation, I don't feel sure enough to argue on either side, and really the whole thing isn't worth arguing about, certainly not here.

I'm certainly going to withhold any predictions until I know the size of M45 and M46 (although theoretically they matter little in comparison to the number of tested exponents to date). Actually since M45 and M46 are the purpose of this thread, maybe we should limit our discussions to them.

[Jeff Gilchrist]

Originally Posted by davieddy

We DO care Tony and how is Jeff getting on?

Sorry, busy day, my verification finished shortly after midnight last night. Verified as prime! So now we have two independent verifications with two different software packages (Mlucas and Glucas) on two different platforms (Sparc and Itanium2). Tony will be finished shortly I'm sure.

Tony: Yes, the system because a little less busy so I was able to speed up a little, but still not much, just enough to stay ahead.

[ewmayer]

Congrats, jeff! So are the lights back on in Canada again, then?