Reply With QuoteI voted Sieving because we are putting a dent in that old sieving range...get a few more people doing it and we can really run thru some range.
Sierpinski problem solved
17 riesel k's remaining (creating SB 2)
Sieving complete until 1000 T (1P)
n upper bound for all remaining k's is above 40M
100 million digit prime discovered
Year 2020
Which of the above will come first?
Any guesses?
I voted Sieving because we are putting a dent in that old sieving range...get a few more people doing it and we can really run thru some range.
definatly could especially if any new advances are made to the algorithim. Sieve speeds could also benefit greatly from more users. After ll isn't only something like 2% of CPU power dedicated to SB used for sieving? Then again maybe it'll never reach 1000T if factor density drops off much more.
I think it will be in this order:
[list=1][*]complete till 40M[*]100M digit prime[*]riesel problem down to 17 k's[*]year 2020[*]sierpinski problem solved[/list=1]
I don't know about the sieving, I don't take part in that. My goal is to find a prime
EatMaDust
Stop Microsoft turning into Big Brother!
http://www.againsttcpa.com
I'm not sure exactly how fast we're moving through the 4 million range but I'm sure we're doing it alot fasster than we did the 3 million range. I'm guessing that this is attributed to either additional sieving or more likely P-1 factoring and moreresources joining our effort. With this in mind I'm gonna say we will finish up to 20 million before the end of 2005 and maybe sooner is a couple primes are found.
Any new thoughts on this. Anyone?
well, GIMPS has maybe just found a new large prime, but I doubt it's 100M
EatMaDust
Stop Microsoft turning into Big Brother!
http://www.againsttcpa.com
There was some hints realeased, so we know that it isn't over 100M digits.
If we continue at current speed in sieving, we should be at 1000T in about two years.
A 100M digit prime might be discovered before then, but I doubt it.
I haven't done any math on this, but to get to 17 remaining Riesels would take a lot of work (atleast 20M, possibly over 40M).
n upper bound. We should be at or near 8M in a year, 40M will take ... a long time.
I think it will happen in this order:
1. Sieving complete until 1000 T (1P)
2. 100 million digit prime discovered
3. 17 riesel k's remaining (creating SB 2)
4. n upper bound for all remaining k's is above 40M
5. Year 2020
6. Sierpinski problem solved
Last edited by ceselb; 11-23-2003 at 01:29 PM.
where is the reisel search right now? Is it progressing much or have they been waiting awhile for their next prime as we have?
A prime was found on 15th october. 99 k's remaining right now (was 101 when it started).
Sieveing is at 2.5T and LLR (PRP) on the k's currently owned by the project is all above 610000 and rising fast (some are much higher).
With 99 k's left they should find a lot more primes than we do.
The lower n's should help 'em, too...Originally posted by ceselb
With 99 k's left they should find a lot more primes than we do.
Hey, I recently stumbled across the prime Sierpinski search. Ain't it possible for them to use some of our tools to speed up their search?
Been there, done that.
I made their dat, finishing two days before Paul Jobling mailed me a version of SoBSieve that could create it automatically. My method included newpgen and some good ol' cut&paste.
Sadly prothsieve doesn't work with their dat as it's hardcoded to num_k=12. This will be a problem for us too when we find the next prime. Talked to mklasson about it and he will do a version without that limitation when he has time.
It's good to see somebody else working on Sierpinski search as well. With n values that low, they sure would crack a few ks soon. Good luck to all.
Any tools that can help this project will be welcomed. A automated PRP client will really help us remove some k.Originally posted by Mystwalker
The lower n's should help 'em, too...
Hey, I recently stumbled across the prime Sierpinski search. Ain't it possible for them to use some of our tools to speed up their search?
Thanks,
Citrix
Citrix,
Did you contact Joe McLean about this?
smh,
Yes I did and we have almost rechecked 70% of the work that was already done. Will be up to 100% by the end of this year.
Citrix
I'm saying 17 Riesel numbers...but I think I'm a little on the biased side
I do realize what a monumentous task it is though...we should be down to 50 k's this time next year with the way our plans for future rate are projecting out as.
BTW...I'm an optimist.
B2
That Riesel Sieve dude
With the help of the people at rieselsieve.com the PSP project has a new forum. The new forum is located at:-
http://www.b2project.com/phpBB2/index.php
Please visit.
Thanks,
Citrix
I see my vote for 17 Riesel numbers is starting to become more of a possibility...we hit another prime yesterday
Lee Stephens
B2
www.rieselsieve.com
I think PSP www.geocities.com/eharsh82/ will be the next SB2, although not on the poll. We have 28 k left for which we need to find a prime.
Citrix
Possible idea:
I don't think that it's wise to include the PSP k's into the SoB.dat (since the sieving depths are so different), but I may be wrong. What's your opinion on that?
Do you mean by PSP k's, the 6 k's that SOB is working on and overlap with our project. If yes, they are not included in the sob.dat. We won't be working on them until SOB releases those k's. We are hoping that SOB will pretty soon find 6 primes for those K's.
Citrix
We won't be working on them until SOB releases those k's.
Well, I additionally meant the 90 other k's...
What 90 other k's. We just have 22 other K's. I think you are confused and should visit the PSP status thread.
http://www.b2project.com/phpBB2/viewforum.php?f=13
Citrix
Right, I meant the other 22.
Mixed it up with the remaining Riesel k's.
So why don't you think it is a good idea?
Citrix
Of course if we'd started sieving at the same time, then obviously it would make sense, but lets look at the realities.So why don't you think it is a good idea?
SoB is at 200T, PSP is at 4T. So for PSP, 4T - 200T needs to be sieved alone from SoB. The 22 remaining candidates could be added to SoB's 11, but that then slows down SoB a lot, and the benefit to PSP right now would be minimal - 200T generates about 2% of what you're get at 4T. So if we have (say) 50 PCs sieving on SoB it would be better to move one onto PSP that to add 22 ks to SoB.
Keep fishing for business, and when the factors run out here you can be sure that sievers will drift to PSP or Riesel or others.
Originally posted by MikeH
Of course if we'd started sieving at the same time, then obviously it would make sense, but lets look at the realities.
SoB is at 200T, PSP is at 4T. So for PSP, 4T - 200T needs to be sieved alone from SoB. The 22 remaining candidates could be added to SoB's 11, but that then slows down SoB a lot, and the benefit to PSP right now would be minimal - 200T generates about 2% of what you're get at 4T. So if we have (say) 50 PCs sieving on SoB it would be better to move one onto PSP that to add 22 ks to SoB.
Keep fishing for business, and when the factors run out here you can be sure that sievers will drift to PSP or Riesel or others.
Mike,
I understand that sieving PSP and SOB k's at the same time is not a good idea right now and I had not requested any one to do so.
But now you raised the point let me explain how things are planned.
We are expecting about 10 -11 K's more to be eliminated under n= 2million. Once done we will go deeper.
We are sieving to a nmax of 2 million only because we have less than 4 sievers at present and if we went to nmax= 20 M we would have been able to sieve very little in the time we had and would have been unable to support the PRP effort simultaneously.
If some one is interested in sieving the PSP and SOB dat's together we can collaborate. I am open to all ideas and suggestions.
We could create a new dat from 2million to 20 million and sieve it to 1T and then you people could sieve it with your k's from 200 T upwards and we can later fill in the gaps.
Once again, if anyone is interested just let me know.
Thanks,
Citrix
I just remembered that sieving (with SoBSieve) was alomst linearly dependent on the number of k's - so that a combined sieve wouldn't improve overall performance a lot (even if Riesel was at 200T). Synergy effects are present when combining n ranges.
IIRC, prothsieve is a bit better, but I'm not sure if there was a big overall performance increase even when noth projects were on the samt sieving depth.
As I've said - all that's just current memorizations - and here, it's 2:36 AM...
Originally posted by Mystwalker
I just remembered that sieving (with SoBSieve) was alomst linearly dependent on the number of k's - so that a combined sieve wouldn't improve overall performance a lot (even if Riesel was at 200T). Synergy effects are present when combining n ranges.
IIRC, prothsieve is a bit better, but I'm not sure if there was a big overall performance increase even when noth projects were on the samt sieving depth.
As I've said - all that's just current memorizations - and here, it's 2:36 AM...
First of some clarifications
When I say SOB.dat I mean the dat file used by PSP and not by SOB even though it is named SOB.dat
Ill ask Mklasson to rename this to proth.dat to avoid confusion.
Leaving this regarding what you posted,
I don't know how Sobsieve works and I can't say anything about it. As for proth_sieve it will take less time to sieve the sob and the PSP k's together than sieving them seperately using 2 machines. (about a 25 % increase).
Citrix
Citrix....you are right...it will be easier to sieve that...however..now you have the task of catching the PSP sieve up to the SoB sieve so that both are at the same level and can be sieved together without one recieving the ranges the other has been thru. Let's see...go from Whatever T you are at..to like 200T...shouldn't take you longLeaving this regarding what you posted,
I don't know how Sobsieve works and I can't say anything about it. As for proth_sieve it will take less time to sieve the sob and the PSP k's together than sieving them seperately using 2 machines. (about a 25 % increase).
What I personally see is the sieve coming to a hault around 250T or so for SoB...and more focus shifted to P-1 testing. Then later...when/If PSP sieves to 250T and has more K's..then you can combine and then sieve them together...however...your two .dat files need to have the same n range....and to my knowledge Citrix you have a max n of about 2 million. If you would set up a psp.dat with n values ranging from 1 million to 20 million and sieve these out to the level of about 250T..then I can see a joint sieving of +1 numbers.
This is the reason I choose to have a n value to 20 million. I believe it is possible that one day we can sieve +1 and -1 numbers at the same time. After several discussions with the sieve guru's..I'm even more sure this can happen...at a future date that may be up to 2 years away. I mean we are just now hitting 10T...we still have 96k's in our riesel.dat...we are still going at a snails crawl that will only be helped out with finding some primes and reducing the number of k's to increase sieve speed. CPU wise we are probably equal to the SoB effort for sieve...but our sieve speeds are 7 times slower..... We could hit 20-30 primes this year and that would really help on sieve speeds.
So..with all that said...I'm afraid we at Riesel Sieve are in a better situation to have a future joint sieve with SoB than you are. The working out of +1 and -1 are trivial compared to you having to go back...make a psp.dat for n=1mil to 20 mil and sieving that up to 250T.
Hopefully none of this even has to happen. Hopefully SoB will bam bam bam hit some primes and the power of CPUs taking out some heavy k's and reducing the work load to pile on the next k will domino effect and the project gets wrapped up in short order....it would be nice...doubtful..but nice. Hopefully PSP can crank out a bunch of primes before you start getting to difficult n values...and maybe our Riesel Numbers will feel the unholy onslaught of power we will throw at it...and decide to give up peacefully before any of them are hurt...doubt it...primes always run....and we will catch them....but it's a thought.
Lee Stephens
B2
www.rieselsieve.com
3 down....96 to go...
Lee, I really do love your optimism. 26 primes = one every two weeksWe could hit 20-30 primes this year and that would really help on sieve speeds.
Good luck.
Assuming there will be a client as user-friendly as the one of SoB (or maybe implmented as a SoB v3 plugin), I really think that's possible.
When there are so many primes to find, it will attract a lot of ppl...
Just to share some good news with you guys:-
PSP is now down to 19k's. We found 3 primes in 2 days. Leaving this we also have a network client working. Hopefully we will find several more primes soon. I think we will be SB2.
Citrix
http://www.geocities.com/eharsh82/
Citrix, congratulations on your primes and good luck for the following primes.
PSP is Seventeenorbust2. 17 k's are left. Just found are 9th prime.
http://www.geocities.com/eharsh82/
Citrix
Only 15 k's remain for PSP now! We found two more primes since the last update.
Citrix
Congratulations!
But remember:
At PSP's n level, SoB also was down to 16 k's left. When it's the same with PSP, you'd find 4 more primes soon...
Well, I guess this hasn't happened...Originally posted by b2uc
What I personally see is the sieve coming to a hault around 250T or so for SoB...and more focus shifted to P-1 testing.
Just realized this part of the posting. Is this really the case? Is the sieving effort not proportional to the k's?Originally posted by Citrix
As for proth_sieve it will take less time to sieve the sob and the PSP k's together than sieving them seperately using 2 machines. (about a 25 % increase).
Last edited by Mystwalker; 05-12-2004 at 06:06 AM.
In theory it will be faster but in reality I don't know. I have never made a real test myself. But it would be better to combine the k's and then sieve than sieving them seperately, it will on the whole increase the speed. Let me know if you want to combine the k's and sieve after 250T.
Citrix
Posting Permissions
