After n=333333, our next exponent will be... (revised poll) - Page 2

jmblazek · 2007-01-16, 17:55

I agree, n=585000 is nice at 3 X 195000. This will give us a 175,000 digit twin.

Also, n=666666 is nice at 2 X 333333. Or 678910 also works 6, 7, 8, 9, 10. These will give us a 200,000 digit twin.

Nevertheless, both of these n's, with current computing power of the project, would take over 9 years to reach 90% probability.

But hey, we got lucky with the first twin...maybe we'll get lucky again. Also, PG was just starting to rev up...so maybe these are doable in a reasonable time frame (2-4 years).

pschoefer · 2007-01-16, 20:13

Quote:

Originally Posted by jmblazek

I agree, n=585000 is nice at 3 X 195000. This will give us a 175,000 digit twin.

Also, n=666666 is nice at 2 X 333333. Or 678910 also works 6, 7, 8, 9, 10. These will give us a 200,000 digit twin.

Nice numbers, but I think we should go with smaller steps, like 333333, 444444, 456789, 543210, 555555, ...

Quote:

Nevertheless, both of these n's, with current computing power of the project, would take over 9 years to reach 90% probability.

But hey, we got lucky with the first twin...maybe we'll get lucky again. Also, PG was just starting to rev up...so maybe these are doable in a reasonable time frame (2-4 years).

In the last search at PrimeGrid, only one quarter of all active participants took part, so if all participants run llr the next time, we can finish it in a acceptable time.

biwema · 2007-01-16, 20:20

My favorite is 500000 or later maybe 666666.

Before making a decision. Check where the fft thresholds are (with a large k (of 1T)). Maybe the point is just few thousands below the decision. Then it could be 50% (guess) faster.

MooMoo2 · 2007-01-16, 22:32

Quote:

Originally Posted by biwema

My favorite is 500000 or later maybe 666666.

Before making a decision. Check where the fft thresholds are (with a large k (of 1T)). Maybe the point is just few thousands below the decision. Then it could be 50% (guess) faster.

I'm leaning toward 500000 too. It's a nice number, near an FFT threshold, and breaks the 150,000 digit barrier. This threshold occurs before 524288 (2^19), so 524288 is not a good number.

pacionet · 2007-01-18, 15:28

But on January 31, we decide the next exponent (and I can start sieving) or we decide the range (and then a new poll start to decide the exponent in that range ???

)

I hope the first one,...

MooMoo2 · 2007-01-18, 16:24

Quote:

Originally Posted by pacionet

But on January 31, we decide the next exponent (and I can start sieving) or we decide the range (and then a new poll start to decide the exponent in that range ???

)

I hope the first one,...

We'll know what the next exponent is by Jan 31st. I'll choose the exponent, which will come from the range that gets the most votes.

pacionet · 2007-01-19, 10:22

For n=500,000 , which seems to be the most probable candidate, what should be the reasonable sieve limit and the probabilities to find a twin ?

biwema · 2007-01-19, 12:02

Quote:

Originally Posted by pacionet

For n=500,000 , which seems to be the most probable candidate, what should be the reasonable sieve limit and the probabilities to find a twin ?

See my realier post in the bonic section:

Quote:

500000
FFT size: 48K
Time 1 test: 500 s
Optimal sieve bits: 57.5
Tests per twin: 23.7M
Twin every G: 90
90% chance after G: 207
CPU years per twin: 376

To have the smae chance of finding a twin I suggest sieve a 250G. The optimal sieve limit will be around 100P to 200P.

Note: After NewPGen is merging the Range at 1G, the save file will be around 4 -4.4 GB. Make sure that your system supports that. Otherwise split the range and merge them after 1T.

jasong · 2007-01-19, 22:17

Quote:

Originally Posted by biwema

See my realier post in the bonic section:

To have the smae chance of finding a twin I suggest sieve a 250G. The optimal sieve limit will be around 100P to 200P.

Note: After NewPGen is merging the Range at 1G, the save file will be around 4 -4.4 GB. Make sure that your system supports that. Otherwise split the range and merge them after 1T.

I'll sieve k=1 to 100G if no one else claims it. I'll go to at least 10^15.

Unless someone already started that?

MooMoo2 · 2007-01-19, 22:40

Quote:

Originally Posted by jasong

I'll sieve k=1 to 100G if no one else claims it. I'll go to at least 10^15.

Unless someone already started that?

Pacionet will be sieving the next n, after 333333.

biwema · 2007-01-20, 20:34

There are different kinds of twin search areas:

1. Interesting or beautiuful exponents: 333333, 500000, 600000, 666666 ...

2. Milestones: 100000 digits, Exponent >=500000, 200000 Digits...

3. Efficient Exponents: Below a FFT limit. (Maybe not to lose, otherwise the number of roundoffsmay increase.

4. Exact number of digits: here the sieving range will be in one decade (for example k between 80G- 800G. Search range is not infinite.

5. other choice

Maybe people can also vote what kind of candidate they want to test.

View Poll Results: The next exponent after n=333333 will be...
under n=400K	2	8.70%
between n=400K and 460K	4	17.39%
between n=460K and 520K	12	52.17%
between n=520K and 580K	1	4.35%
above n=580K	4	17.39%
Voters: 23. You may not vote on this poll

2007-01-18, 15:28	#16
pacionet Oct 2005 Italy 153₁₆ Posts	But on January 31, we decide the next exponent (and I can start sieving) or we decide the range (and then a new poll start to decide the exponent in that range ??? ) I hope the first one,... Last fiddled with by pacionet on 2007-01-18 at 15:29

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Stockfish game, move 11 poll. Not "Trump vs Clinton" poll.	MooMoo2	Other Chess Games	0	2016-11-07 06:07
n=333333 Twin Found!	Kosmaj	Twin Prime Search	12	2009-08-14 13:10
What if we don't find twin prime n=333333?	cipher	Twin Prime Search	5	2009-04-16 21:53
n=333333 off the list	TimSorbet	No Prime Left Behind	2	2008-05-13 17:26
Preparing N=333333	biwema	Twin Prime Search	26	2007-02-11 23:35

2007-01-16, 17:55	#12
jmblazek Nov 2006 Earth 100₈ Posts	I agree, n=585000 is nice at 3 X 195000. This will give us a 175,000 digit twin. Also, n=666666 is nice at 2 X 333333. Or 678910 also works 6, 7, 8, 9, 10. These will give us a 200,000 digit twin. Nevertheless, both of these n's, with current computing power of the project, would take over 9 years to reach 90% probability. But hey, we got lucky with the first twin...maybe we'll get lucky again. Also, PG was just starting to rev up...so maybe these are doable in a reasonable time frame (2-4 years).

2007-01-16, 20:20	#14
biwema Mar 2004 381₁₀ Posts	My favorite is 500000 or later maybe 666666. Before making a decision. Check where the fft thresholds are (with a large k (of 1T)). Maybe the point is just few thousands below the decision. Then it could be 50% (guess) faster.

2007-01-19, 10:22	#18
pacionet Oct 2005 Italy 3×113 Posts	For n=500,000 , which seems to be the most probable candidate, what should be the reasonable sieve limit and the probabilities to find a twin ?

2007-01-20, 20:34	#22
biwema Mar 2004 3·127 Posts	There are different kinds of twin search areas: 1. Interesting or beautiuful exponents: 333333, 500000, 600000, 666666 ... 2. Milestones: 100000 digits, Exponent >=500000, 200000 Digits... 3. Efficient Exponents: Below a FFT limit. (Maybe not to lose, otherwise the number of roundoffsmay increase. 4. Exact number of digits: here the sieving range will be in one decade (for example k between 80G- 800G. Search range is not infinite. 5. other choice Maybe people can also vote what kind of candidate they want to test.