[XC] riptide

07-15-2006, 03:58 AM

http://www.seventeenorbust.com/download/

This is the download page for your client.

http://www.seventeenorbust.com/faq/

This is the FAQ page. In short this is a Maths project trying to find the smallest Sierpinski numbers

Extract from the wiki:

"A problem in NumberTheory proposed by Waclaw Sierpinski (1882-1969). First, a definition:

Consider numbers of the form N = k * 2^n + 1, where k is odd and n > 0. If, for some fixed k, every integer n yields a composite (non-prime) number N, then k is said to be a Sierpinski number.

The SierpinskiProblem, simply put, is: What is the smallest Sierpinski number?

JohnSelfridge proved in 1962 that k = 78557 is a Sierpinski number. The proof shows that every choice of n falls into at least one of seven categories, where each category guarantees a factor of N. A detailed explanation of the proof is available on TeamPrimeRib's web site.

Most mathematicians believe that 78557 is, indeed, the smallest Sierpinski number. But to prove it, we have to show that every number k < 78557 is not a Sierpinski number. Remember, a Sierpinski number is a fixed k such that all n yield composite N. So a non-Sierpinski number is a fixed k such that some choice of n yields a prime N. This turns out to be relatively easy to do for most choices of k. However, sometimes n has to grow very large before a prime number appears.

By early 2002, primes had been found for all but seventeen choices of k. At that point, the SeventeenOrBust project began a systematic DistributedComputing search of the remaining k values. Nine of them have now been eliminated by the project, and eight remain to prove that 78557 is the smallest Sierpinski number. The community is divided on the question of whether or not it is likely the SeventeenOrBust project will complete its search within its authors' lifetimes. Heuristics have been used to estimate the range of numbers that must be tested before eliminating all the remaining multipliers is likely, but most of these heuristics have been demonstrated to be inaccurate. In any case, it is very likely that SeventeenOrBust will be able to eliminate at least some of the remaining eight.

The remaining k candidates are 10223, 19249, 21181, 22699, 24737, 33661, 55459 and 67607. "

Where are we going?

Yesterday (14/07/2006) we were 17th in overall production with 2.747TceMs.

Yesterdays Top producer TeamPrimeRib was 32.896TceMs. (T for Terra)

Overall (15/07/2006) we are 84th but with approx 10 people returning work we are still owning our neighbours so we will rise fast.

Overall TeamPrimeib have 18.680PcEMs (Peta)

Guess who's in 2nd? Oh hell yes!!! Its the DPC with about half this at 9.88P

http://stats.free-dc.org/new/allteams.php?proj=sob

My 2 cents...

CPU order of rank in SOB.

1)Core - Conroe and later (not mobile Core architecture)

2)Netbursts (with 512K cache or over), Cellys may not do as well as may be expected (even at >4GHZ)because of 256K cache. If you do have one, tighten up memory settings.

3)Amds. AMD's do OK but in a DC community could be used to much better effect on another project. But they ain't no slouches. Just that they compare to Netbursts at the SAME clockspeed.

Quick instructions

1 )http://www.seventeenorbust.com/download/

2)http://www.seventeenorbust.com/signup/

Create an account.

3)PASSWORD WILL BE SENT TO YOU VIA EMAIL

4)Login into site and change your affiliation to team 'XtremeSystems' in preferences.

5)Go to your client and THEN hit the signup button. Enter in required information. Press start. For all subsequent instances you just have to enter name and team under config button on client.

6) Enjoy!

NOTE: Our password is xs (lowercase)

When you install a second client, you don't need to 'create a new login'. Just fill your teamname & username in the fields on the config page.

Also I think that if you uncheck 'Transmit intermediate blocks' it may help performance as the client momentarily stops crunching while communicating.

NOTE: IT IS WELL WORTH READIN THE README'S THAT COME WITH THE CLIENT. THEY PROVIDE NICE INFORMATION FOR SERVICE INSTALL AND COMMAND SWITCHES ETC. ALSO SHOW HOW TO USE HT TO ITS BEST EFFECT. ONE MUST BE CAREFUL ALSO TO SET SOB ON A PRIORITY THAT WILL ALLOW OTHER CPU DC PROJECTS TO SHARE CPU WITH IT AHOULD YOU NEED TO.

Well nothing more to add.... but I think we're doing quite well with our limited membership and invite others of all CPU persuasions to join up.

Update: if your on dial up or have another reason to queue your work, you can always get SBQueue http://home.att.net/~k.brazier/programs/java/sbqueue/index.html#download For windows download the keepqueuelevel.zip on the link and this has scripts that are handy to follow.

Update: running 3 clients on a DX is slightly (about 5%) better than 4 clients. Use 'sobsvc -o2' as a service install for this. See more in the readme's.

UPDATE. Important!! If you receive a block and you do not report intermediate results or just simply stop crunching it within 30 days that block is discarded and any credits you have earned on the intermediate uploads will be lost and taken back

If you drop a test ie. if you restart SOB and it does not finish the test, but downloads another new one you will lose all interim credits after 30days. There is a fix however. Check this (http://www.xtremesystems.org/forums/showthread.php?t=113852)thread first. The link to the fix is in it

More updates: Do you ever notice the 'explorer.exe' process in task manager using more proc. power than it should? Say 5-8 %? Apparently thats a bug in the service install setup. If you adjust the Seventeen or Bust service to not 'interact with desktop' this proc usage will go down to normal levels for 'explorer.exe'

This is the download page for your client.

http://www.seventeenorbust.com/faq/

This is the FAQ page. In short this is a Maths project trying to find the smallest Sierpinski numbers

Extract from the wiki:

"A problem in NumberTheory proposed by Waclaw Sierpinski (1882-1969). First, a definition:

Consider numbers of the form N = k * 2^n + 1, where k is odd and n > 0. If, for some fixed k, every integer n yields a composite (non-prime) number N, then k is said to be a Sierpinski number.

The SierpinskiProblem, simply put, is: What is the smallest Sierpinski number?

JohnSelfridge proved in 1962 that k = 78557 is a Sierpinski number. The proof shows that every choice of n falls into at least one of seven categories, where each category guarantees a factor of N. A detailed explanation of the proof is available on TeamPrimeRib's web site.

Most mathematicians believe that 78557 is, indeed, the smallest Sierpinski number. But to prove it, we have to show that every number k < 78557 is not a Sierpinski number. Remember, a Sierpinski number is a fixed k such that all n yield composite N. So a non-Sierpinski number is a fixed k such that some choice of n yields a prime N. This turns out to be relatively easy to do for most choices of k. However, sometimes n has to grow very large before a prime number appears.

By early 2002, primes had been found for all but seventeen choices of k. At that point, the SeventeenOrBust project began a systematic DistributedComputing search of the remaining k values. Nine of them have now been eliminated by the project, and eight remain to prove that 78557 is the smallest Sierpinski number. The community is divided on the question of whether or not it is likely the SeventeenOrBust project will complete its search within its authors' lifetimes. Heuristics have been used to estimate the range of numbers that must be tested before eliminating all the remaining multipliers is likely, but most of these heuristics have been demonstrated to be inaccurate. In any case, it is very likely that SeventeenOrBust will be able to eliminate at least some of the remaining eight.

The remaining k candidates are 10223, 19249, 21181, 22699, 24737, 33661, 55459 and 67607. "

Where are we going?

Yesterday (14/07/2006) we were 17th in overall production with 2.747TceMs.

Yesterdays Top producer TeamPrimeRib was 32.896TceMs. (T for Terra)

Overall (15/07/2006) we are 84th but with approx 10 people returning work we are still owning our neighbours so we will rise fast.

Overall TeamPrimeib have 18.680PcEMs (Peta)

Guess who's in 2nd? Oh hell yes!!! Its the DPC with about half this at 9.88P

http://stats.free-dc.org/new/allteams.php?proj=sob

My 2 cents...

CPU order of rank in SOB.

1)Core - Conroe and later (not mobile Core architecture)

2)Netbursts (with 512K cache or over), Cellys may not do as well as may be expected (even at >4GHZ)because of 256K cache. If you do have one, tighten up memory settings.

3)Amds. AMD's do OK but in a DC community could be used to much better effect on another project. But they ain't no slouches. Just that they compare to Netbursts at the SAME clockspeed.

Quick instructions

1 )http://www.seventeenorbust.com/download/

2)http://www.seventeenorbust.com/signup/

Create an account.

3)PASSWORD WILL BE SENT TO YOU VIA EMAIL

4)Login into site and change your affiliation to team 'XtremeSystems' in preferences.

5)Go to your client and THEN hit the signup button. Enter in required information. Press start. For all subsequent instances you just have to enter name and team under config button on client.

6) Enjoy!

NOTE: Our password is xs (lowercase)

When you install a second client, you don't need to 'create a new login'. Just fill your teamname & username in the fields on the config page.

Also I think that if you uncheck 'Transmit intermediate blocks' it may help performance as the client momentarily stops crunching while communicating.

NOTE: IT IS WELL WORTH READIN THE README'S THAT COME WITH THE CLIENT. THEY PROVIDE NICE INFORMATION FOR SERVICE INSTALL AND COMMAND SWITCHES ETC. ALSO SHOW HOW TO USE HT TO ITS BEST EFFECT. ONE MUST BE CAREFUL ALSO TO SET SOB ON A PRIORITY THAT WILL ALLOW OTHER CPU DC PROJECTS TO SHARE CPU WITH IT AHOULD YOU NEED TO.

Well nothing more to add.... but I think we're doing quite well with our limited membership and invite others of all CPU persuasions to join up.

Update: if your on dial up or have another reason to queue your work, you can always get SBQueue http://home.att.net/~k.brazier/programs/java/sbqueue/index.html#download For windows download the keepqueuelevel.zip on the link and this has scripts that are handy to follow.

Update: running 3 clients on a DX is slightly (about 5%) better than 4 clients. Use 'sobsvc -o2' as a service install for this. See more in the readme's.

UPDATE. Important!! If you receive a block and you do not report intermediate results or just simply stop crunching it within 30 days that block is discarded and any credits you have earned on the intermediate uploads will be lost and taken back

If you drop a test ie. if you restart SOB and it does not finish the test, but downloads another new one you will lose all interim credits after 30days. There is a fix however. Check this (http://www.xtremesystems.org/forums/showthread.php?t=113852)thread first. The link to the fix is in it

More updates: Do you ever notice the 'explorer.exe' process in task manager using more proc. power than it should? Say 5-8 %? Apparently thats a bug in the service install setup. If you adjust the Seventeen or Bust service to not 'interact with desktop' this proc usage will go down to normal levels for 'explorer.exe'