Processing  2^811-1 is continuing but it is an extremely large and difficult computation, so we will not be seeing factors for some time yet.   More news when we have it.

December 2nd 2003
On December 2nd, 2003, NFSNET completed the factorization of 2^757-1. The C213 composite cofactor was found to be the product of primes with 79 and 134 decimal digits. See the announcement for more details. Thanks to all who participated in this project and keep up the sieving on 2^811-1!

November 1st 2003
The filtering for 2^757-1 is now complete and the linear algebra stage started on October 28th. Filtering reduced the 76.6 million relations to matrix with about 7.5 million rows and columns. Fifteen of the sixteen dual-processor nodes of the cluster at MSRC are working on the matrix and current estimates predict a runtime of about 30 days.

October 13th 2003
We've completed the sieving stage for 2^757-1 and the relations are being gathered at Cambridge for the post processing. About 24.5 million lines were sieved and over 76.5 million relations were collected. We expect the post processing will take about 3 weeks to complete.

Now it's on to 2^811-1 and potentially a record-breaking SNFS factorization!

October 10th 2003
NFSNET has almost finished the sieving stage for 2^757-1. Based on our current estimates, sometime on Monday we will have collected enough relations and will switch to the next project. The relations will be gathered for post-processing and, barring any unforseen problems, we should know the factors of 2^757-1 in a few weeks time.

Our next project is 2^811-1, a Mersenne number. 2^811-1 is a composite number with 245 decimal digits. It has just one known factor, 326023. The remaining composite cofactor has 239 decimal digits. Enough ECM effort has been done on the cofactor to be fairly certain that it has no factors of 50 digits or less.

To date the largest number factored using SNFS is 2^809-1, factored by Franke/Kleinjung/Bahr in early 2003. Assuming we have no challengers (never a safe assumption!) the factorization of 2^811-1 will establish a new worldwide SNFS record.

This project will be much more difficult than the previous projects undertaken by NFSNET. As with 2^757-1 we'll need about 80 million relations but the work required to find them will be substantially more, about six times more.

With that in mind, we need as many contributing users and machines as possible. With our current level of participation the sieving stage could take as long as one year but we will be starting various recruitment activities to attract more participation.

So fire up your machines and spread the word. This is your chance to participate in a record-breaking factorization.

September 15th 2003
Sieving on 2^757-1 continues. As of midnight September 13th we have collected just over 58 million relations of the approximately 78 million relations that we expect we'll need.

Later this year we intend to start the 239-digit cofactor of 2^811-1, also known as M811. Before then we have to ensure that enough ECM work has been done to give a good probability that no factors with under 50 digits remain to be found. This work is in progress and until it completes, NFSNET will continue to factor Most Wanted Numbers from the Cunningham tables. The next in line is 5,302+.c187

September 1st 2003
Despite delays caused by Paul Leyland going on vacation for two weeks, the factors of 10^227-1 and 2^713-1 both completed without mishap. Full details may be found on the announcments page.

August 20th 2003
The cluster at MSRC has completed the linear algebra for 10^227-1 however it unfortunately did not complete until a few hours after Paul Leyland left on vacation. When he is back at the beginning of September the square root stage will be started and hopefully the factors of 10^227-1 will be revealed shortly afterwards.

Immediately after that the post-processing for 2^713-1 will continue and unless there are any unforseen problems should complete in one to two weeks.

Meanwhile the sieving on 2^757-1 continues. As of midnight August 19th we have collected almost 24.4 million relations of the approximately 78 million relations that we expect we'll need.

July 24th 2003
The sieving for 10_227M, aka 10,227-.c212, finished on 7 July some 70 days after it began. Since then a lot has been happening behind the scenes. This is by far the largest and hardest factorization NFSNET has yet attempted and we expected to run into new problems.

Starting on 8th July and running overnight, 1.8 gigabytes of compressed files were pulled over from Richard Wackerbarth's systems in Austin to my workstation in Cambridge. Uncompressing the data expanded it to close to 4Gb. There were 74,973,761 relations in the complete set.

Singleton removal reduced the set to 56.9M relations held in four files for a total of 3.5Gb. Further filtering steps reduced the data to 39.45M relations in 8.76M sets. The relations file was split into 40 pieces and each piece fully factored in turn, using a P4 workstation and 20 cpu nodes on the MSRC cluster.

A test matrix was build and it had 8,555,397 rows and 8,924,965 columns which is over-square by 369,568. A last filter run was started and it discarded 365049 heavy relation sets. The matrix built from the output has 7,925,462 rows and 7,930,179 colums and so is oversquare by only 4717.

The linear algebra began at 10:25 this morning. It is running on 15 out of the 16 nodes in the MSR cluster (the omission being one node with suspected broken hardware). As each node has 2 cpus, there are 30 processes running, one per cpu.

One process is the master and it is using 240M of active virtual memory; the 29 slave processes are using 99M each, for a total of something over 3Gb. Each machine has 2Gb, so there is more than enough RAM to go around. The nodes communicate with each other over gigabit ethernet through a dedicated switch. Each node is using around 20% of its available bandwidth, so the switch is passing something like 3Gbps on average with peaks perhaps twice that. The program has been running for 5 hours now, which is long enough to be able to estimate that the complete run will take close to 445 hours --- barring interruptions.

July 23rd 2003
A version of the NFSNET client for UltraSparcs running Solaris is now available. See the Join section to download this new version.

July 9th 2003
The sieving for 10^227-1 finished on July 7th. 10^227-1 was substantially more difficult than the previous projects. Total calendar time was about 70 days and during this time there were about 85 machines active for most of the period. Over 17.6 million lines were sieved producing about 74.8 millions relations.

Originally it was anticipated that as many as 82 million relations might be required so we were quite pleased to see that just under 75 million turned out to be sufficient.

The relations are now at Cambridge and post-processing has begun. Unless any unforseen problems are encountered, the post-processing phase should be complete in two to three weeks.

The 2^713-1 project has been underway for a few days now. Assuming that participation levels remain about the same as they are now, the sieving for 2^713-1 should be complete near the end of July. We estimate that we'll need 43.5 million relations before the post-processing can begin.

July 6th 2003
The project statistics have been updated to include a new measurement called 'work units'. Until now participation has been measured in terms of lines sieved or relations collected, however, neither of these is a good metric to judge the degree of work performed by participants.

The number of lines sieved is not a good measure because projects are often divided into multiple subprojects and the subprojects can have lines of different length. Also, different projects may use line lengths of very different sizes. The number of relations collected per line varies depending on various properties of the line including whether it is near the start or end of the sieve area and whether or not the line number is divisible by many small primes.

The introduction of the work unit metric provides a way to measure the contribution of participants in a way that is consistent across projects and subprojects. The work unit calculation takes into account the line length and factorbase size of projects and subprojects to arrive at a scaled metric that closely matches the work actually being done by the siever.

June 12th 2003
As of yesterday we have sieved about 10.5 million lines and collected about 44.2 million relations for 10^227-1.

We estimate that we'll need slightly less than 80 million relations before we can start the post-processing steps so as you can see we still have a good distance to go.

Our combined processing power has been growing steadily. We're now sieving approximately 325K lines per day whereas one month ago we were closer to an average of 225K lines per day. The number of users participating daily has remained fairly steady in the range of 35-40 but the number of machines active from those users has grown from the high 80s to over 100 with the record at 107 on June 10th.

May 20th 2003
NFSNET has completed the factorization of 12^197-1. On May 20th, it was found that the C212 co-factor is the product 70- and 143-digit primes. For more details see the announcement.

May 13th 2003
The linear algebra for 12^197-1 started yesterday. Filtering produced a matrix with 5,144,636 rows and 5,168,041 columns. We estimate that the linear algebra phase will take about 174 hours to complete, or about 7.25 days. As before, the cluster at MSRC is being used. Fifteen cluster nodes with a total of thirty 1Ghz P-III CPUs are being used. The current jobs are taking 66MB per process, 132MB per node, except for the single master process which is using 157MB.

May 12th 2003
Release Candidate 1 (RC1) of the NFSNET client is now available for download. RC1 is the first version that has been made available to the general public without requiring a request to join the testing program. To download and install the NFSNET client, visit the join page and follow the instructions there.

May 12th 2003
Release Candidate 1 (RC1) of the NFSNET client is now available for download. RC1 is the first version that has been made available to the general public without requiring a request to join the testing program. To download and install the NFSNET client, visit the join page and follow the instructions there.

May 1st 2003
Sieving has completed for 12^197-1. The relations are now being collected and the remaining post-processing steps are being started.

The sievers are now getting assignments for 10^227-1. With a 212-digit composite cofactor, 10^227-1 has the potential to yield a record-sized penultimate factor.

Apr 29th 2003
The sieving for 12^197-1 is nearing completion. Approximately 40 million of the 43.5 million required relations have been found and the NFSNET sievers will be switched to a new project when the 43.5 million mark is reached.

NFSNET has selected 10^227-1 for its next factorization. This project will be significantly more difficult than previous ones and will exercise the NFSNET data handling abilities at higher levels than before. 10^227-1 has a composite cofactor with 212 digits and has the potential to yield a record-size penultimate factor.

Apr 26th 2003
NFSNET has completed the factorization of 5^298+1. The 189 composite cofactor was split into primes of 50 and 139 digits. See the announcement for additional details.

Apr 17th 2003
The filtering phase for 5^298+1 is complete and the linear algebra will be starting shortly. The matrix to be processed has 4,897,158 rows and 4,903,379 columns.

In order to reduce the weight of the matrix and decrease the run-time of this phase, the primes < 120 have been excluded. These small primes accounted for a weight of 40,881,211. With them removed the weight of the remaining matrix is 247,805,504.

Apr 14th 2003
The relations for 5^298+1 have now arrived in Cambridge. This time the relations were sent by FTP and the transfer required approximately 9 hours. At total of 43,578,824 relations were collected including 5,336 duplicates.

The current version of the NFSNET client software includes a few new features that simplify the administration of the collection of clients. In particular, it includes a redirection feature which was recently tested successfully. Clients can be redirected from server to server and this was used to redirect clients while the Texas server was inaccessible due to network reconfiguration at that site. From now on, clients will be moved from server to server as necessary, for instance to perform load balancing.

Mar 29th 2003
The sieving stage for 5^298+1 has completed. The relations are now being collected and will undergo initial filtering in Texas before they are sent to Cambridge for post-postprocessing.

We've have now switched to sieving on 12^197-1. As with the previous project, we'll need about 42 million relations before post-processing can begin. This project will be a bit more difficult than the previous one. Trial sieving suggests that we'll need to sieve up to at least b=11M.

Mar 20th 2003
NFSNET has completed the factorization of 12^178+1. On March 20th, it was found that the C145 co-factor is the product 69- and 76-digit primes. See the announcement for complete details.

Mar 17th 2003
NFSNET has completed the factorization of 2^673-1. On March 17th, it was found that the C151 co-factor is the product 59- and 92-digit primes. See the announcement for complete details.

Feb 25th 2003
NFSNET has completed the factorization of 11^197+1. On Feb 25th, it was found that the C161 co-factor is the product 76- and 85-digit primes:

P76 = 9892817292333841349326992570525012475558072147134506608232615478 \
288870172703

P85 = 1840014589329963784595624375113478453521208836735048403925274125 \
097969736250506946877

Feb 24th 2003
A small subset of sievers have been switched to the new 5^298+1 project while the majority are finishing off the remaining sieving for 12^178+1. Very shortly we will have enough relations to start the post-processing for 12^178+1 and all sievers will be switched to 5^298+1.

The post-processing for 11^197+1 is nearing completion and we hope to announce the results in a few days. When it is complete we'll start the post-processing for 2^673-1 and it will be followed by 12^178+1.

Feb 17th 2003
The sieving stage for 2^673-1 has completed and the relations are now being collected for the post-processing steps. We sieved approximately 6.5M lines and found over 43M relations.

The switch has been thrown to activate our next project, 12^178+1, and the NFSNET collection of sievers is now working on it. 12^178+1 is another of the Most Wanted Numbers on the Cunningham list and, assuming successful completion, will be the fourth Cunningham MWN that NFSNET has tackled.

Filtering for 11^197+1 has been under way for a few days now and the linear algebra stage should start soon. Once that is complete the post-processing for 2^673-1 will begin.

Feb 12th 2003
Sieving continues on 2^673-1. So far we have sieved over 4.6M lines and collected about 34.5M relations. We estimate that we will require an additional 8M relations before the post processing phase can be started

Feb 4th 2003
NFSNET has completed the factoriztion of a second Cunningham Most Wanted Number.

The factors of the C156 cofactor of 5^289+1 are:

p62 = 12156847076444678217726556516258597180670422099432269314205601 and
p95 = 10693062703280568569932417392486626220178668772845830558089170\          255922593409573680162209144318227

Jan 30th 2003
Linear algebra started on 5^289+1. It's likely to take a week or so to complete.

Jan 29th 2003
Factorization of 6^257-1 discovered. The factors have 58 and 115 digits.

Sieving completed for 11^197+1 completed and a new project started for 2^673-1.

Jan 17th 2003
Participation statistics for 11^197+1 have been posted and will be updated daily from now on. Also, in the report of individual computers you can now drill into a particular computer and see its day-by-day contribution. We'll be improving the reports over time, adding more reports, graphs and other enhancements.

Also, the sieve data for 6^257-1 and 5^289+1 have arrived in Cambridge and the post processing steps started on Jan. 14th. Linear algebra for 6^257-1 will start next week and is expected to require about 5 days if all 32 processors on the cluster are available. Once that is finished, the post processing for 5^289+1 will start and it is expected to take about the same amount of time.

Jan 9th 2003
We've posted basic statistics for the 5^289+1 sieving effort. Use the link on the righthand side of this page to view the statistics. There are a few inaccurracies and we don't have complete identification for all of the contributing machines. We'll improving on this as time goes on and hope soon to have daily updates of current progress.

If you were running the NFSNET client software during the period of December 7th, 2002 to January 4th, 2003 and your user id is not listed above, please send mail to don@nfsnet.org.

Jan 4th 2003
The sieving for 5^289+1 is now complete and we are now working on 11^197+1, another of the most wanted numbers on the Cunningham list.