Estimate of sieve duration for 2^811 - 1

dsouza123 · 2003-10-28, 04:58

I read the news at nfsnet.org mentioning that 2^811 could take as long as year.

http://www.nfsnet.org/status.html

I did some calculations, using the current rate of production the sieving would finish in July 2004.

There are roughly 70 machines ( 35 users ) active each day.

If you aren't currently working on nfsnet, consider joining in.

Wacky · 2003-10-28, 06:55

Thanks for encouraging others to join. We NEED a lot more help.

First the good news. Paul has his cluster busy on the post-processing of M757 and other things. We should be able to get some time from it before we finish this sieving for M811.

But for the bad news. I'm not sure what method you used to make your estimate. If you are extrapolating relations per day, you must recognize that we are presently sieving at relatively low line numbers. The number of relations per line will drop as we get further into the work, but the effort will remain the same.

I seriously doubt that we will meet your projection unless there is a massive increase in participation.

But we WILL get there........

eventually.

Richard

dsouza123 · 2003-10-28, 21:13

I didn't realize that the number of relations per line sieved would vary substantially, it is about 2.25 relations per line now, so could it get as low as 1 for 1 or even lower ?

If it requires alot of work to figure it out please don't spend your time on it.

I hope more people will join in.
At least having this spot in the mersenneforum gives very good exposure.

Thanks to all who make it possible.

I would never have known about it otherwise.

Dan

Wacky · 2003-10-28, 22:48

2.25 is the number that I hae for lines around 4M. The MSRC server jumped back to fill in 2M and is getting about 3.4 there.
Up at 10M, the LoneStar folks are getting about 1.63.

Above 50M, I don't have good numbers yet, but I am estimating that we will be below 0.75. Depending on how things fill in, we may have to go even higher and get only 0.65.

As we go along, I'll be able to improve my estimates. But, in any case, it's a lot of sieving. Bring your friends. There is plenty of work to spread around.

Richard

junky · 2004-01-23, 02:27

hi, im new into that forum. Im running distributed contest for a couple of years now (RC5-64, SETI@home, ECCp-109, ECC2-109 (in progress) and im testing NSF now :)
i dont know if ya were expecting these numbers:
but we have:
40.69% complete (32553.3 of 80000 workunits)

later

junky · 2004-02-27, 01:03

hhehe, now the % is over 85%, pretty cool and we're still in February.

and for today, we reached a new record of lines ( http://www.nfsnet.org/stats2/statsre...gesize=&page=0 ) . so its really cool.

btw, i've a question for Richard, ya said: "2.25 is the number that I hae for lines around 4M. The MSRC server jumped back to fill in 2M and is getting about 3.4 there.
Up at 10M, the LoneStar folks are getting about 1.63."

whats the MSRC server and LoneStar is the name of a pool right ?

thanks.

dsouza123 · 2004-02-27, 01:11

MSRC = Microsoft Research Cambridge
Lonestar is the Texas server run/maintained by Wacky ( Richard ).

junky · 2004-02-27, 01:31

Thanks for clarification :)

can i know who's is maintening the Fenland, and Longhorn server ?

frmky · 2004-02-27, 01:54

Quote:

Originally Posted by junky

and for today, we reached a new record of lines ( http://www.nfsnet.org/stats2/statsre...gesize=&page=0 ) . so its really cool.

Actually the stats are a little funky right now. The total number of relations has dropped since yesterday. But I'm sure all will be back to normal in short order. Move along. Nothing to see here.

Greg

dsouza123 · 2004-02-27, 02:03

Longhorn is the new name for Lonestar, so it is maintained by Richard.

As for Fenland, I don't know for sure, but Fenland is near Cambridge so it probably is Paul.

junky · 2004-02-27, 02:14

thanks dan, sorry greg

hhehee

2003-10-28, 04:58	#1
dsouza123 Sep 2002 2×331 Posts	Estimate of sieve duration for 2^811 - 1 I read the news at nfsnet.org mentioning that 2^811 could take as long as year. http://www.nfsnet.org/status.html I did some calculations, using the current rate of production the sieving would finish in July 2004. There are roughly 70 machines ( 35 users ) active each day. If you aren't currently working on nfsnet, consider joining in.

2004-02-27, 01:03	#6
junky Jan 2004 133₁₀ Posts	hhehe, now the % is over 85%, pretty cool and we're still in February. and for today, we reached a new record of lines ( http://www.nfsnet.org/stats2/statsre...gesize=&page=0 ) . so its really cool. btw, i've a question for Richard, ya said: "2.25 is the number that I hae for lines around 4M. The MSRC server jumped back to fill in 2M and is getting about 3.4 there. Up at 10M, the LoneStar folks are getting about 1.63." whats the MSRC server and LoneStar is the name of a pool right ? thanks. Last fiddled with by junky on 2004-02-27 at 01:07

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
How many ecm curves per B1 is a good estimate?	EdH	Factoring	45	2013-10-23 13:39
Benchmark Estimate	Primeinator	Information & Answers	8	2009-06-11 23:39
V25.7 TF estimate way out ... or am I?	petrw1	PrimeNet	5	2008-11-08 02:23
Estimate the new primes	davieddy	Lounge	34	2008-09-17 04:22
A tricky way to estimate pi(x)	XYYXF	Math	13	2006-09-01 15:44

2003-10-28, 06:55	#2
Wacky Jun 2003 The Texas Hill Country 2101₈ Posts	Thanks for encouraging others to join. We NEED a lot more help. First the good news. Paul has his cluster busy on the post-processing of M757 and other things. We should be able to get some time from it before we finish this sieving for M811. But for the bad news. I'm not sure what method you used to make your estimate. If you are extrapolating relations per day, you must recognize that we are presently sieving at relatively low line numbers. The number of relations per line will drop as we get further into the work, but the effort will remain the same. I seriously doubt that we will meet your projection unless there is a massive increase in participation. But we WILL get there........ eventually. Richard

2003-10-28, 21:13	#3
dsouza123 Sep 2002 296₁₆ Posts	I didn't realize that the number of relations per line sieved would vary substantially, it is about 2.25 relations per line now, so could it get as low as 1 for 1 or even lower ? If it requires alot of work to figure it out please don't spend your time on it. I hope more people will join in. At least having this spot in the mersenneforum gives very good exposure. Thanks to all who make it possible. I would never have known about it otherwise. Dan

2003-10-28, 22:48	#4
Wacky Jun 2003 The Texas Hill Country 3²×11² Posts	2.25 is the number that I hae for lines around 4M. The MSRC server jumped back to fill in 2M and is getting about 3.4 there. Up at 10M, the LoneStar folks are getting about 1.63. Above 50M, I don't have good numbers yet, but I am estimating that we will be below 0.75. Depending on how things fill in, we may have to go even higher and get only 0.65. As we go along, I'll be able to improve my estimates. But, in any case, it's a lot of sieving. Bring your friends. There is plenty of work to spread around. Richard

2004-01-23, 02:27	#5
junky Jan 2004 10000101₂ Posts	hi, im new into that forum. Im running distributed contest for a couple of years now (RC5-64, SETI@home, ECCp-109, ECC2-109 (in progress) and im testing NSF now :) i dont know if ya were expecting these numbers: but we have: 40.69% complete (32553.3 of 80000 workunits) later

2004-02-27, 01:11	#7
dsouza123 Sep 2002 2×331 Posts	MSRC = Microsoft Research Cambridge Lonestar is the Texas server run/maintained by Wacky ( Richard ).

2004-02-27, 01:31	#8
junky Jan 2004 7×19 Posts	Thanks for clarification :) can i know who's is maintening the Fenland, and Longhorn server ?

2004-02-27, 02:03	#10
dsouza123 Sep 2002 2×331 Posts	Longhorn is the new name for Lonestar, so it is maintained by Richard. As for Fenland, I don't know for sure, but Fenland is near Cambridge so it probably is Paul.

2004-02-27, 02:14	#11
junky Jan 2004 205₈ Posts	thanks dan, sorry greg hhehee