Twin prime search?

[MooooMoo]

Originally Posted by Joshua2

What percent chance that we will have found it by Sept 07? Is it around 50-50?

Assuming PG continues at the current rate of ~1.6G per month, we'll be at 14.5 G by the end of September. This means that the chance of finding a twin by Sept 07 will be 30%.

For a 50-50 chance, we'll have to LLR about 29G, which will happen sometime around during the summer of 2008.

[Terminal]

Originally Posted by MooooMoo

Assuming PG continues at the current rate of ~1.6G per month, we'll be at 14.5 G by the end of September. This means that the chance of finding a twin by Sept 07 will be 30%.

For a 50-50 chance, we'll have to LLR about 29G, which will happen sometime around during the summer of 2008.

But, theorically, with the new server, PG speed could increase: less tech stops, less "server load too high" stops.

We are doing roughly 1TFlop now, we could may be go up to 2-3 TFlop with the new server!

Edit: or somebody else might find a twin over 333333 and we have to jump directly into 500000 (or other!)

[MooooMoo]

Originally Posted by MooooMoo

Estimated day we'll be the second highest project in terms of number of primes found (assuming the current rate continues): May 24

TPS and PG are at #2 today, which is three weeks earlier than I predicted. Good job guys!

http://primes.utm.edu/bios/top20.php...&by=PrimesRank

[Terminal]

Just a copy from my post at PG: it gives some projected dates for our study for n=333333 taking into account PG progress.

just for number lovers... weekly update

Status at 17/05/07 - 17:15 my time... today!

Checking range at 7955 M
Posibility to have found a twin by then was: 17.86%
Obviously we did not (or may be is on the way as part of the range is still WIP!)

Range per day- we have accelerated a LOT!! -: 176,85M/day: then...

50% prob point (28G aprox): 07/09/2007 (7th September)
75% prob point (56G aprox): 12/02/2008 (12th Feb next year)
end of analysis (100G point): 18/10/2008 (18th Oct next year...)
By that day, prob to have found twin is 91,5% roughly.

1 prime found every 3689 WU or,
1 prime found every 10,9M range k's (roughly)

Addtionally:
By 50% prob point, whe would have found 2564 titan primes
by 75% prob point, 5128,
by the end of this exponent, (100G point), 9157...

unfortunatelly, only about 1600 titan primes at 333333 exponent will ever be on the top 5000 list: when we discover more, the olders will drop from the list.

Good news: Increase in speed is GREAT! we did not achieve 180M/d because of the resent units yesterday & today and the short stop Sunday morning.

Proposal to Rytis & TPS: we should get - after 9200M range already in the loop at PG - the new sieved file for our work here. I think we are still using the old one, not reflecting the k´s that are being cleared out from the sieving effort. We should get a 5-10% speed increase just by that!

See you next week!

[MooMoo2]

So this was the thread that started the project. The first post was made on April 13, but was edited a month later. I guess I'm in a nostalgic mood, so I'm wondering how the original unedited post looked like before those edits. Maybe Oddball can bring it up.

Looking back on things, TPS has been through quite a ride. There were two record twins, a flood of primes on the top 5000 list, Primegrid coming and leaving, and a new moderator stepping up. There's also a strange feeling of being at the beginning again - there are only a few participants, individual primes are too small to be on the top 5000 list, the project leader is relatively new to the forum, and gribozavr is doing almost all of the sieving. The project started less than five years ago in 2006, but that year seems to be an age away...

BTW, I'd like to try some LLR tests for the variable n range, but there's no file that has the k/n pairs with the p<600T factors removed

[Oddball]

Originally Posted by MooMoo2

So this was the thread that started the project. The first post was made on April 13, but was edited a month later. I guess I'm in a nostalgic mood, so I'm wondering how the original unedited post looked like before those edits. Maybe Oddball can bring it up.

Can't do that. I can bring up the original posts for the sieve and LLR reservation threads, but the original post for this thread was edited too long ago.

The project started less than five years ago in 2006, but that year seems to be an age away...

It sure does! The economy was booming back then and single core processors dominated, among other things.

BTW, I'd like to try some LLR tests for the variable n range, but there's no file that has the k/n pairs with the p<600T factors removed

I think gribozavr has the file on a machine that is currently down:

http://www.mersenneforum.org/showpos...&postcount=229

I'm waiting for it too, because we're looking to add LLRnet and PRPnet support for TPS.