Riesel Sieve Boinc

[SlangNRox]

I just wanted to try and drum up support for rieselsieve boinc. We are currently in 29th place and should get up to 28th in the next couple of days. Currently there is just 2 people on the team working on this project (me and aka 1nas) who has about 45% of the total teams score. So if you would like to help find a new big prime number sign up. So far 1 prime has been found under boinc and it weighed in at 296,000 digits. (Alot more have been found using their older non boinc client)

So if you like to join me and help TA advance start crunching some rieselsieve wu under boinc. It should be listed under "attach to project" or enter in "http://boinc.rieselsieve.com/" when prompted. It is important to go into your preferences page and select run test projects so you can try to find primes. This is better suited for intel chips.

If anyone has any questions I"ll try to answer them.

BTW there has been discussion about rieselsieve granting more credit per cpu time than other projects. So for the time being running this may help you get more overall boinc pts on some of those combined boinc worldwide stats pages.

 

[Fardringle]

My main home PC is attached to every active BOINC project (and a bunch of inactive ones ). It gives each of the projects a turn now and then so Riesel Sieve does get some points from me once in a while but it's really not the type of project that interests me enough to take any of my other dedicated machines off of the projects they are running.

However, due to problems with Rosetta not playing nice on my server lately, I just might be convinced to put Riesel Sieve on it to help the team for a while if the client behaves properly in the SBS 2003 operating system.

 

[SlangNRox]

I believe the client behaves really well. I know there was a problem with the linux sieve so they moved all the linux clients to llr which is better since thats what will find the primes.



I'd also like to welcome MK.
That makes 4 of us here crunching.

 

[Fardringle]

It hasn't been running long yet, but so far the server seems happy with Riesel Sieve. Four copies of the client are only using a bit under 100MB of RAM, which is just slightly better than the 3GB+ that Rosetta was taking.

This server was getting about 7.5 points per hour per core (four instances on two Xeon CPUs) on Rosetta. It still had some work queued up from Rosetta and Leiden when I enabled Riesel Sieve on Wednesday so it hasn't turned in many results yet but the seven results it has returned so far got 14.7 points per hour per core from the RS client. That's double what it was getting from Rosetta, by the way.

If those point averages stay the same once it goes "full time" on Riesel Sieve it should be turning in roughly 1400 points per day. That's not too shabby for a two-year-old machine. I'm sure it would be much better if it was a Core 2 Quad server instead of a P4 Xeon system, but I don't think I can convince the boss to upgrade just so that I can get more BOINC stats.

 

[mk]

Thanks for the welcome, SlangNRox.

 

[SlangNRox]

I guess David Anderson and some other boinc people are complaing to rieselsieve about the amount of points that they are giving.
There is a meeting planned for tomorrow to discuss this.

I'm glad to see some other people crunching along with me.

 

[networkman]

Well, I just re-attached to the project with my work machine, although the share is set so that only 1/4 of the cycles go to RS -- that'll change when the Enigma project takes a nap for a few weeks.

And actually, I'd forgotten that I already had an account set up with a fair amount of completed work, so there will be a re-awakening of my account.

 

[Fardringle]

After one week of crunching nothing but Riesel Sieve, it looks like my server is putting out an average of 1050 points per day. That's less than the 1400 it would get if it was doing nothing but crunching, but since it is actually used as a server for about 9 hours each day, I'll take what I can get.

 

[SlangNRox]

Welcome back networkman. YOu just blew blew right past past me in daily output.

 

[networkman]

Originally posted by: SlangNRox
Welcome back networkman. YOu just blew blew right past past me in daily output.

How so? I look Here and it shows you clearly with a much higher RAC than I.

 

[Fardringle]

I wouldn't pay much attention to the ASA on the project web site. I have been doing about 1200 points per day for the past week but my ASA on that link is only 623. It looks like Riesel Sieve uses the average over the past 14 days to calculate its ASA, so we'll have to wait two weeks before it shows an accurate representation of what we're actually doing.

 

[SlangNRox]

Welcome Danzigrules to the rs project.


Don't forgot to go to the riesel sieve boinc page click on "your account" and under the rs pref select run test projects. That will be a great help the project by testing for primes instead of sieving and eliminating candidates that have factors and are not prime. It allows you the chance to find a prime and the sieving has progressed so far, running test projects is much better. The test are longer but you get around 115 pts per test now. My 2.4ghz quad takes under 4 hrs per test per core to run them.

 

[Fardringle]

I'm jealous. My P4 Xeons take about 10 hours per test when running 4 at a time. It's a dual processor server running 4 threads (HT CPUs). I'm going to have it just run 2 processes at a time for a day or two to see what happens.

 

[danzigrules]

thanks,

I did the test project thing too

 

[networkman]

I changed my prefs to allow for the test apps too.

 

[Fardringle]

Originally posted by: Fardringle
I'm jealous. My P4 Xeons take about 10 hours per test when running 4 at a time. It's a dual processor server running 4 threads (HT CPUs). I'm going to have it just run 2 processes at a time for a day or two to see what happens.

Just over 6 hours per test when running one thread per processor. About 10 hours per test when running two per processor so I guess I'll stick with four at a time.

 

[networkman]

Originally posted by: Fardringle

Originally posted by: Fardringle
I'm jealous. My P4 Xeons take about 10 hours per test when running 4 at a time. It's a dual processor server running 4 threads (HT CPUs). I'm going to have it just run 2 processes at a time for a day or two to see what happens.

Just over 6 hours per test when running one thread per processor. About 10 hours per test when running two per processor so I guess I'll stick with four at a time.

Ah, so that's what those LLRs are - tests! Okay.. I was wondering, 'cuz one of 'em took just about 10.5 hours to complete! :Q

 

[Fardringle]

They take longer to run than the normal work units, but they're worth more points.

 

[SlangNRox]

Welcome danzigrules and anyone else that I have misseed. We had 10 member return results in today and jumped 2 spots to #25. Amazing output by everyone.

 

[networkman]

Congratulations to danzigrules in passing me so handily. I may be able to increase output just a little bit more, but not near enough to provide any real competition for ya.