mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Conjectures 'R Us (https://www.mersenneforum.org/forumdisplay.php?f=81)
-   -   All top 5000 primes will have the same number of digits in 2014 ;-) (https://www.mersenneforum.org/showthread.php?t=18039)

Batalov 2013-03-26 23:18

All top 5000 primes will have the same number of digits in 2014 ;-)
 
I remember the times when 150,000-digit primes were enough to enter...

Now, this 283939-digit prime is only going to stay for a week in the top-5k. :max:

gd_barnes 2013-03-27 10:24

When I started in prime searching in 2007, the big push was to get all of the annoying base 2 n=333333 (~100k digits) primes knocked out of the top 5000. I personally still find fixed n searches annoying. The n=1000000 and n=1290000 primes clog up the top 5000 and knock off "good" primes. They also skew the long term curves of the rising 5000th place prime.

Puzzle-Peter 2013-03-27 14:42

Fixed-n is just about the only way to go if you're looking for twins, sophies, triples etc. The candidates grow much too quickly otherwise.

Mini-Geek 2013-03-27 15:21

[QUOTE=Puzzle-Peter;335092]Fixed-n is just about the only way to go if you're looking for twins, sophies, triples etc. The candidates grow much too quickly otherwise.[/QUOTE]

You could vary n, you just need to search a large k range. Consider if you were to search n=1.0M-1.1M. You might lose some sieving efficiency from not having one fixed n, but you'd gain some efficiency by being able to search smaller k's. I don't know which effect is larger.

Puzzle-Peter 2013-03-27 17:27

[QUOTE=Mini-Geek;335097]You could vary n, you just need to search a large k range. Consider if you were to search n=1.0M-1.1M. You might lose some sieving efficiency from not having one fixed n, but you'd gain some efficiency by being able to search smaller k's. I don't know which effect is larger.[/QUOTE]

Might be worth a try. In fact I have already done this for quad-hunting, without taking timings though. But would a bunch of 1.0M to 1.1M primes be much better than bunch of 1.0M primes? How can primes be "good" or "bad" anyway?

gd_barnes 2013-03-28 04:39

[QUOTE=Puzzle-Peter;335109]Might be worth a try. In fact I have already done this for quad-hunting, without taking timings though. But would a bunch of 1.0M to 1.1M primes be much better than bunch of 1.0M primes? How can primes be "good" or "bad" anyway?[/QUOTE]

It would be just as bad. The issue is: Twins and SGs should not be searched for in the top 5000 ranges. It takes too long to find them and ruins the graph of the rising 5000th, 1000, 500th, etc. place primes. IMHO, twins should be searched at n=800K right now.

All of the n=1290000 primes demotivate smaller searchers because the 5000th place prime rises so rapidly. Soon you will need 50+ cores to have a reasonable chance to find a top 5000 prime within a few weeks. The small guy is quickly being phased out. I guess prime searching is a microcosm of the real world.

Puzzle-Peter 2013-03-28 14:47

Thanks Gary, now I understand what's bothering you. It's a difficult question. The large BOINC twin or Sophie projects need a lot of small guys working together but many of these small guys would like to have the chance of getting a single prime into the TOP5000 even if the twin test gives a negative result. I don't think many of them would be hunting primes on their own or in a non-BOINC project as we do here.

The bar is raising, that's for sure. On my hunt for triples and quads I am looking at CPU-centuries when aiming for the top and I even have to worry about proving the numbers prime. I am now in the process of proving a number on which primo failed on the first attempt. Fortunately Marcel Martin told me about an entry in the .ini file that seems to have done the trick although it's not finished yet... is there anybody keeping a list of PRP twins, triples, quads etc that are impossible to prove atm?

MyDogBuster 2013-03-28 22:25

[QUOTE]I don't think many of them would be hunting primes on their own or in a non-BOINC project as we do here. [/QUOTE]It's a shame what has happened to the Top5000. It used to be a nice goal to attain. Now it's just a joke. Kinda takes the fun right out of finding primes.

Maybe we should start our own Top100 (think small) for just NPLB & CRUS.

kar_bon 2013-03-28 23:27

The anoying point with PrimeGrid is the fact, that, although there're many participants and so many different prime searches, PG is still unable to make a simple list of primes they've found, like on my pages or like [url=http://www.prothsearch.net/riesel.html]ProthSearch[/url]. For search ranges you have to go to their forum, greater primes can be found on the Top5000 page, non Top5000 primes you have to look at their PRPnet resultfiles. Even members seem to have problems, where their found primes go.
I've mentioned this in their forum long long time ago but nothing happend.
PG is more in working with challenges and making points for members in different classes instead of making the results available for others in a handsome manner.

MyDogBuster 2013-03-29 00:06

I used to be a BOINCer until I realized how juvenile the points system is. Any single project can distort their awards and make any measurement mute.

It is a real shame that the last thing PRIME Grid wants to do is actually make finding the prime important.

Keep up the good work Karsten. There are people who enjoy your database.

rogue 2013-03-29 01:29

[QUOTE=MyDogBuster;335285]I used to be a BOINCer until I realized how juvenile the points system is. Any single project can distort their awards and make any measurement mute.[/QUOTE]

I agree with that. BOINC will punish or reward people based upon their hardware. In other words, if you have a specific hardware feature BOINC can increase or decrease your credit regardless of throughput.


All times are UTC. The time now is 04:21.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.