20050221, 11:20  #1 
5×1,823 Posts 
BeoWolf Clusters
Does anyone know anything about running Prime95 on a BeoWolf cluster?

20050221, 11:38  #2 
3·5·283 Posts 
BeoWolf = Beowulf
Sorry about the spelling! BeoWolf=Beowulf
The reason for the question is my friend and I were discussing really 'heating up' an overclocked processor and Intel's site said Prime95. That lead to your site and our personal 'thread' of thought then became crunching primes. The next logical semi LOW CO$T step could possibly be a BeoWulf cluster... We tech support computers and quite often an owner will give us their old equipment. This has led to several tall stacks of old machines P1  200Mhz and above we can't bear to throw out yet. If we procured some fast fibre cards... perhaps a BeoWulf cluster? The question is... would Prime95 scale to a BeoWulf cluster and has anyone done this? There is probably a thread already on this but we are 'newbies' to your fine forum and haven't had much chance to look around. We do apologize in advance if this is already addressed. Thank you for your time. 
20050221, 11:54  #3 
May 2004
Oslo, Norway
170_{8} Posts 
The Linux version of Prime95 is named mprime, and is probably the version you would want to run on a Beowulf cluster.
The question has been raised here before, and I think the consensus is that it's not worth the bother. As the GIMPS software already is a distributed project, you're far better off just setting up a copy of the program on each of the computers and let them chug away, each on their own exponent. Any clustering technology will introduce its own overhead, thus slowing down the overall process. One other point worth mentioning: Unless you live in the cooler parts of the world, and need to heat up the house anyway, running GIMPS on a heap of antiquated hardware is IMO a waste of energy. regards, Leif. 
20050221, 17:43  #4 
Jul 2004
Nowhere
329_{16} Posts 
i can offer some help
http://www.beowulf.org/overview/howto.html info site http://www.phy.duke.edu/~rgb/Beowulf...ook/index.html good book/still in devolopment on web http://www.g4tv.com/screensavers/fea..._Clusters.html something done by techtv on baywolfs comes of of deep hole of links\/ http://ganglia.sourceforge.net/ http://wwwunix.mcs.anl.gov/mpi/ could be usefil http://www.csm.ornl.gov/pvm/pvm_home.html also could be usefil http://bofh.be/clusterknoppix/ premade cluster version of bootable knoppix prime 95 donest run on a cluster envrioment Last fiddled with by moo on 20050221 at 17:44 
20050222, 15:40  #5 
5·1,297 Posts 
Thank you all!
Thank you for the comments.
We'll eventually put more machines on Prime95 and overclock them. Hopefully we won't melt to many uProcessors. Hehee !!! If we do we'll get some pics and link them here. We finally got Prime95 to work after several hours of "error 23 Unable to communicate with server," It's now 44.22% This sure beats SETI for system idle time. The chances of finding anything were quite remote with SETI. The chances of finding a mersenne prime are almost 100%. Not that we would be THE ONE to find it, but we can help sift out the final prime from the chaff. Prime95 was responsible for what, Something like 5 of the last mersenne prime finds? Does anyone have a reasonalby accurate guess as to what Mxx Prime95 should hit the 10 million digit prime? Is there a mathmatical formula to predict that? If it does exist is it linear? Glad to be onboard! Thanks again !!! 
20050222, 22:31  #6  
6809 > 6502
"""""""""""""""""""
Aug 2003
101×103 Posts
5·7·311 Posts 
Quote:
Quote:


20050223, 04:51  #7 
1111000111111_{2} Posts 
Nice Link
That was a great link:
Well the last prime found and the one being verified are seeming to mess with the trend a bit, but check this link: http://www.utm.edu/research/primes/...xtMersenne.html It is LINEAR!!! I was thinking that this prime business is probably one of the most interesting things since Fibonacci. Then from the link above were found all kinds of Fibonacci Prime stuff. Whoohooo!!! Math heaven!!! There may be a typo or something at: http://primes.utm.edu/glossary/page....FibonacciPrime within the last sequence of numbers, shown here with <> around numbers that are prime forward and reversed, [] around number forward not prime factors, and {} around number reversed not prime factors: <3>, 4[2.2], <5>, <7>, 9[3.3], 14[2.7], <17>, 21[3.7]{2.2.3}, 25[5.5]{2.2.13}, 26[2.13]{2.31}, 65[5.13]{2.2.2.7}, 98[2.7.7], 175[5.5.7], <191>, 382[2.191], 497[7.71]{2.397}, 653{2.2.89}, 1577[19.83]{23.337}, 1942[2.971]{47.53}, 1958[2.11.89]{11.11.71}, 2405[5.13.37]{2.2521}, 4246[2.11.193]{2.2.2.11.73}, 4878[2.3.3.271]{2.2.2.2.3.3.61}, 5367[3.1789]{3.5.509} Alot of these numbers are not Fibonacci sequence numbers anyway. Wanted to get some consensual validation before submitting typos to the author... AM I MISSING WHAT THE AUTHOR IS TRYING TO CONVEY? Sometimes I screw up that way... The Prime link is a VERY GREAT LINK still ! Thank you for your time. 
20050223, 05:10  #8 
1100010000011_{2} Posts 
DIREGUARD PREVIOUS POST !!! I DID SCREW UP
I SCREWED UP, MISSED THE U SUB N... AREN'T SMART GIRLFRIENDS WONDERFUL !!! SHE FOUND IT FOR ME.
DAMN !!! IT'S ALWAYS THE SIMPLE THINGS !!! WHIMPER WHIMPER WHIMPER... = FACE GETTING RED = 
20050223, 05:20  #9 
Jul 2004
Nowhere
809 Posts 
mrnetguy may i sugest you sign up for the forums insted of posting as a guest ;) just a thot

Thread Tools  
Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
Twin Prime Days, Prime Day Clusters  cuBerBruce  Puzzles  3  20141201 18:15 
Clusters! In! Space!  CRGreathouse  Information & Answers  29  20110502 04:33 
P95 and Linux clusters  revivalfire  Information & Answers  3  20080718 16:06 
Beowolf cluster on the Cheap, breaking 100$/GFlop  jflin  Hardware  8  20070906 08:25 
Using PCIE based network hardware for faster math clusters  Peter Nelson  Hardware  0  20050707 00:12 