Can Old PCs Help New PCs on Same Team?

2010-06-12, 04:44

Hi,

A few months ago, a friend recommended the Prime95 program to stress-test the CPU on an old Windows 98 PC. When I looked up the program, I discovered this project and it sounds really cool. Now I'd like to learn a bit more about how the project works. I've looked around the website, but haven't found some of the things I'm looking for, so here I am!

I read on the website about "teams" that participate in the search. At the same time, I understand that slower computers are not ideally equipped to perform the lengthy operations that qualify for the available prizes, BUT that there are secondary operations that can be assigned to these older machines.

Therefore, I would like to know the following: I have two current (Vista) computers and three really old PCs (Pentium, Pentium II, Pentium III). Is it possible to set up a "team" where the older computers will somehow be directly helping the new machines to win the prize?

I won't deny the fact that the cash prize is a big motivating factor here, but I'd like to add that given the odds of actually finding the next large prime number, and the cost of running PCs 24/7, if money were the only factor I'd probably be better off playing the lottery. As I said, this IS a really neat project. It's the combination of the goal plus the prize that got me here.

So then, can I set up my old PCs to provide direct assistance to the new PCs in pursuit of the prize by performing the auxiliary operations, or would the old PCs' efforts be going into some general pool of results?

If this topic has already been covered, I apologize for any annoyance I've caused, and will appreciate getting pointed to the page with the answer.

Thanks very much.

Rodrigo

S485122 · 2010-06-12, 10:13

Quote:

Originally Posted by Unregistered

So then, can I set up my old PCs to provide direct assistance to the new PCs in pursuit of the prize by performing the auxiliary operations, or would the old PCs' efforts be going into some general pool of results?

The old PCs will only contribute to the general pool. Their contribution to your direct search of a 100 000 000 digit prime (to earn the EFF cash price) would be marginal and difficult to arrange...

Jacob

cheesehead · 2010-06-12, 22:36

Rodrigo,

Jacob's correct, but I want to elaborate.

Short answer: Your slow computers are relatively useless in pursuit of the money prize.

Medium answer: Theoretically, it's possible to design software to divide and assign the subtasks of the money prize pursuit in a way that might allow multiple computers of any speed to contribute, each in proportion to its speed. However, that's not the way the existing software was designed.

"Teams" mean something different -- each member and computer operates independently, but their achievement totals are added together for competitive ranking purposes.

Longer answer:

The work coordinated by GIMPS includes several different types of tasks. Only one type (first-time Lucas-Lehmer tests) is used to pursue the money prize, and that type is best performed on fast systems.

All the other work types are valuable from a mathematical point of view. Some of them might earn you a teeny-tiny bit of fame among mathematicians (i.e., your name could be added to a long list somewhere), but most will simply anonymously contribute to group efforts. They can be great fun for those who love math, but will not earn you any money.

If you're still interested (and are prepared for an occasional forum comment from folks who place excessive emphasis on speed), your "slow" systems can usefully perform some of those secondary tasks.

2010-06-12, 22:39

O.K., I understand. I had visions of taking the old computers out of mothballs to somehow help the new ones, but I guess that's not the way the project works.

Thanks for clarifying that!

Uncwilly · 2010-06-13, 00:41

Quote:

Originally Posted by cheesehead

Longer answer:

The work coordinated by GIMPS includes several different types of tasks. Only one type (first-time Lucas-Lehmer tests) is used to pursue the money prize, and that type is best performed on fast systems.

To get real nit picky:
2 types of work are done on numbers before the first-time Lucas-Lehmer test.
They are Trial Factoring and P-1 factoring.

Trial Factoring can be broken up easily among multiple machines. If you chose to try for the prize money for the first 100 million digit prime, you could get an assignment for an exponent. Likely there would be some TF and P-1 work left to do. You could set the older Pentiums to do some/all of the TF while the faster machine does the P-1 (or if the P-III has enough memory, have it do that while the fast machine does starts the L-L, but this could be a waste). Then after the various bits of factoring are done have the fast machine do the L-L. When that happens, you can pick up another number. Have the slow machines work on the factoring while the fast machine works for a year or so on the first number.

2010-06-13, 05:59

cheesehead,

Now that there's more than one answer, I'd better start addressing my posts to avoid confusion.

Thank you for elaborating on what Jacob said. From reading around the website I did think that the way you explained it was the way the projects works, but I wanted to ask just in case my idea was doable. (And Uncwilly's later reply gives hope.)

This is indeed a very interesting project, so if my slow computers actually ended up bogging things down instead of helping, I wouldn't want to use them. Or give grounds for complaints. I don't know for sure yet how things are structured, but in principle I'd think that even a little bit of help from a slow PC should be better than no help from that old PC. Like having an old soldier plugging a hole in the defenses instead of no one.

I appreciate your taking the time to explain!

Rodrigo

2010-06-13, 06:12

Uncwilly,

This is what I was hoping for!

So you're saying that my slower computers could in fact be assigned to assist my faster ones, by doing some of the simpler prep work for the same prime number? That would be really exciting!

I'm going back into the project site to explore the requirements for each type of task, and which versions of the software would work with my respective computers. Hopefully all the PCs I mentioned will be able to contribute. (I also have two ancient Intel 8086/8088 machines, but that's probably pushing it...)

Thanks very much.

Rodrigo

lycorn · 2010-06-13, 18:14

Quote:

Originally Posted by Unregistered

(...) even a little bit of help from a slow PC should be better than no help from that old PC.

You are absolutely right, and that is the best way of approaching the GIMPS project!

Your old PCs (except the 8086/8088) are perfectly usable for Trial Factoring large exponents. You may use them the way Uncwilly described, or you might as well just choose the type of work TF-LMH, that will assign large exponents to your machines, that are returned to the pool once factored to the default bit level.

You are either way very welcome aboard.
Suggestion: register in the forum.
Just curious: where are you from? Rodrigo is a portuguese name, although also found in other countries.

Uncwilly · 2010-06-14, 01:40

It would take a Pentium about 6 months to take a number in the 100 million digit range from 72 to 73. This one would be best used by GIMPS in the TF-LMH type of work (where it would be looking at lower bit levels.)
The other machines could more practically do the trial factoring while the fast machine does P-1, which should take about 1 month (IIRC). If you keep an eye on the machines you can stop them if one finds a factor.

But, dividing up the primality test, L-L to separate CPU's is not worth doing.

There is lots of factoring work needed in the prize money area. (However, the money only gets distributed to those that did the L-L test.)

chalsall · 2010-06-14, 15:04

Quote:

Originally Posted by Uncwilly

It would take a Pentium about 6 months to take a number in the 100 million digit range from 72 to 73. This one would be best used by GIMPS in the TF-LMH type of work (where it would be looking at lower bit levels.)

To expand on this a bit -- at the rate the TF-LMH (and independent TFers) are chewing through the exponents, I predict there won't be any work left in the 63 -> 64 ranges within the next two to three months.

As has been noted elsewhere, older machines lose their effectiveness doing TF work above 64 (i.e. 64 -> 65 or above).

Based on this, it would actually be more cost effective for you (after a period of time) to buy a new machine, taking into consideration work achieved per KWh and your cost per KWh.

2010-06-15, 01:47

lycorn,

Yes, if it's possible I would like to use my old computers to assist my new computers in their pursuit of the prize, as Uncwilly described.

And to answer your question -- I'm not Portuguese, but (if it counts) my grandfather was born in Galicia! :-)

The weekend was very busy so I didn't get to explore the GIMPS website very well, but I do want to get up to speed on L-L, TF, P-1 and the rest of the project vocabulary. Then I will feel more comfortable participating in the forum.

Thank you!

2010-06-12, 04:44	#1
Unregistered 2×3²×109 Posts	Can Old PCs Help New PCs on Same Team? Hi, A few months ago, a friend recommended the Prime95 program to stress-test the CPU on an old Windows 98 PC. When I looked up the program, I discovered this project and it sounds really cool. Now I'd like to learn a bit more about how the project works. I've looked around the website, but haven't found some of the things I'm looking for, so here I am! I read on the website about "teams" that participate in the search. At the same time, I understand that slower computers are not ideally equipped to perform the lengthy operations that qualify for the available prizes, BUT that there are secondary operations that can be assigned to these older machines. Therefore, I would like to know the following: I have two current (Vista) computers and three really old PCs (Pentium, Pentium II, Pentium III). Is it possible to set up a "team" where the older computers will somehow be directly helping the new machines to win the prize? I won't deny the fact that the cash prize is a big motivating factor here, but I'd like to add that given the odds of actually finding the next large prime number, and the cost of running PCs 24/7, if money were the only factor I'd probably be better off playing the lottery. As I said, this IS a really neat project. It's the combination of the goal plus the prize that got me here. So then, can I set up my old PCs to provide direct assistance to the new PCs in pursuit of the prize by performing the auxiliary operations, or would the old PCs' efforts be going into some general pool of results? If this topic has already been covered, I apologize for any annoyance I've caused, and will appreciate getting pointed to the page with the answer. Thanks very much. Rodrigo

2010-06-12, 22:36	#3
cheesehead "Richard B. Woods" Aug 2002 Wisconsin USA 2²·3·641 Posts	Rodrigo, Jacob's correct, but I want to elaborate. Short answer: Your slow computers are relatively useless in pursuit of the money prize. Medium answer: Theoretically, it's possible to design software to divide and assign the subtasks of the money prize pursuit in a way that might allow multiple computers of any speed to contribute, each in proportion to its speed. However, that's not the way the existing software was designed. "Teams" mean something different -- each member and computer operates independently, but their achievement totals are added together for competitive ranking purposes. Longer answer: The work coordinated by GIMPS includes several different types of tasks. Only one type (first-time Lucas-Lehmer tests) is used to pursue the money prize, and that type is best performed on fast systems. All the other work types are valuable from a mathematical point of view. Some of them might earn you a teeny-tiny bit of fame among mathematicians (i.e., your name could be added to a long list somewhere), but most will simply anonymously contribute to group efforts. They can be great fun for those who love math, but will not earn you any money. If you're still interested (and are prepared for an occasional forum comment from folks who place excessive emphasis on speed), your "slow" systems can usefully perform some of those secondary tasks. Last fiddled with by cheesehead on 2010-06-12 at 22:50

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Team name	Brian-E	Game 3 - ♚♛♝♞♜♟ - Morphy's Maniacs	7	2014-12-21 06:16
Team Spirit	davieddy	Lounge	7	2011-10-10 06:39
Team PrimeX Factoring Team	louis_net	Teams	9	2004-10-15 22:59
Where's the Team Stats?	Death	NFSNET Discussion	9	2004-06-12 19:22
Team stats	Prime95	PrimeNet	15	2003-11-24 15:56

2010-06-12, 22:39	#4
Unregistered 20411₈ Posts	O.K., I understand. I had visions of taking the old computers out of mothballs to somehow help the new ones, but I guess that's not the way the project works. Thanks for clarifying that!

2010-06-13, 05:59	#6
Unregistered 2×3²×13 Posts	cheesehead, Now that there's more than one answer, I'd better start addressing my posts to avoid confusion. Thank you for elaborating on what Jacob said. From reading around the website I did think that the way you explained it was the way the projects works, but I wanted to ask just in case my idea was doable. (And Uncwilly's later reply gives hope.) This is indeed a very interesting project, so if my slow computers actually ended up bogging things down instead of helping, I wouldn't want to use them. Or give grounds for complaints. I don't know for sure yet how things are structured, but in principle I'd think that even a little bit of help from a slow PC should be better than no help from that old PC. Like having an old soldier plugging a hole in the defenses instead of no one. I appreciate your taking the time to explain! Rodrigo

2010-06-13, 06:12	#7
Unregistered 14105₈ Posts	Uncwilly, This is what I was hoping for! So you're saying that my slower computers could in fact be assigned to assist my faster ones, by doing some of the simpler prep work for the same prime number? That would be really exciting! I'm going back into the project site to explore the requirements for each type of task, and which versions of the software would work with my respective computers. Hopefully all the PCs I mentioned will be able to contribute. (I also have two ancient Intel 8086/8088 machines, but that's probably pushing it...) Thanks very much. Rodrigo

2010-06-14, 01:40	#9
Uncwilly 6809 > 6502 """"""""""""""""""" Aug 2003 101×103 Posts 2·19·293 Posts	It would take a Pentium about 6 months to take a number in the 100 million digit range from 72 to 73. This one would be best used by GIMPS in the TF-LMH type of work (where it would be looking at lower bit levels.) The other machines could more practically do the trial factoring while the fast machine does P-1, which should take about 1 month (IIRC). If you keep an eye on the machines you can stop them if one finds a factor. But, dividing up the primality test, L-L to separate CPU's is not worth doing. There is lots of factoring work needed in the prize money area. (However, the money only gets distributed to those that did the L-L test.)

2010-06-15, 01:47	#11
Unregistered 1,553 Posts	lycorn, Yes, if it's possible I would like to use my old computers to assist my new computers in their pursuit of the prize, as Uncwilly described. And to answer your question -- I'm not Portuguese, but (if it counts) my grandfather was born in Galicia! :-) The weekend was very busy so I didn't get to explore the GIMPS website very well, but I do want to get up to speed on L-L, TF, P-1 and the rest of the project vocabulary. Then I will feel more comfortable participating in the forum. Thank you!