Can Old PCs Help New PCs on Same Team?

[Rodrigo]

chalsall,

O.K., you’ve won me over. I’d like to take a modified approach to the original idea, and use the two faster machines to “go for the gold” while the three slower ones make general contributions to the GIMPS.

I appeal to you, Uncwilly, cheesehead, NBtarheel, lfm, and Jacob for guidance as to what I can do to accomplish both of those things.

As part of the process of discovery, I visited the Manual Assignments page http://www.mersenne.org/manual_assignment/ but that one introduced (yet) another new term that I hadn’t come across, the “world record tests.” What is that, and how does it fit in with the rest of the project?

One other question about these “manual” assignments: Obviously I haven’t signed up any of my PCs yet, so I’m not clear on how tasks get assigned. Do you get an “automatic” assignment when you first register, or do you have to go that page regardless when you sign up, or is that a place where you can go to get something else that you prefer instead? (I gather that you have the choice to have the server or software make a selection for you, or else you can go to that page if you want something else. But I’m making sure.)

Changing tack now – if I registered all five machines, would I be right to believe that they would all count toward the totals given in the Producers page http://www.mersenne.org/report_top_500/ . In which case there’d be no need to form a team to show up on the Teams page http://www.mersenne.org/report_top_teams/ . Would there be any purpose to forming a team out of one’s various PCs?

I hope that I haven’t tried your patience (or anybody else’s here) by asking all these elementary questions. If nothing else, maybe this thread will help other newcomers who are considering joining GIMPS.

Thank you very much.

Rodrigo

[Rodrigo]

Uncwilly,

Thank you very much for the detailed answers. Slowly my understanding is increasing. And you’ll see that I took your suggestion to register. J

I don’t think that my P75 would want to wait 18 years to finish an operation. Is there any type of work that such a machine could complete in a matter of months, whether alone or in tandem with other slower PCs?

I appreciate your taking the time to help me understand the basics of the GIMPS project. I’m sure there will be a lot more issues as I get into it, but you and everyone else have been very patient with my newbie questions, and for that I am grateful.

Rodrigo

[Uncwilly]

Originally Posted by Rodrigo

the “world record tests.” What is that, and how does it fit in with the rest of the project?

That is a test where, the number is larger than the currently largest known prime number. Currently 2^43112609 -1 is the largest known prime number. If you get an assignment that is a "world record test", it will be bigger than that. Currently, almost all of the "first time tests" numbers are WRT's.

One other question about these “manual” assignments: Obviously I haven’t signed up any of my PCs yet, so I’m not clear on how tasks get assigned. Do you get an “automatic” assignment when you first register, or do you have to go that page regardless when you sign up, or is that a place where you can go to get something else that you prefer instead?

It is generally easiest to do in the software on the machines. For the slower machines choose "Trial factoring to low levels". For the faster machines to try for the prize, choose "100,000,000 digit numbers to test".

In which case there’d be no need to form a team to show up on the Teams page. Would there be any purpose to forming a team out of one’s various PCs?

Teams are best if you have multiple people working together. No need to form a team.

Go poke around in the Lone Mersenne Hunters forum:
http://www.mersenneforum.org/forumdisplay.php?f=12
You can learn over there how to set the older machines to work in the high range already talked about. Like NBTarheel suggests. A P90 machine will take ~20 hour to take one of those numbers from 62 to 63 bits.

Originally Posted by Rodrigo

I don’t think that my P75 would want to wait 18 years to finish an operation. Is there any type of work that such a machine could complete in a matter of months, whether alone or in tandem with other slower PCs?

See the last part of my response.

Most all of us have been newbies at one point in time.

[chalsall]

Originally Posted by Uncwilly

You can learn over there how to set the older machines to work in the high range already talked about. Like NBTarheel suggests. A P90 machine will take ~20 hour to take one of those numbers from 62 to 63 bits.

But keep in mind there are *no* exponents at below 63 "bits" other than very low exponents.

Rodrigo, I would really encourage you to not even bother bringing your P75 online. But if you must (and for your other slower machines) may I suggest you work in the 889M range? It's not currently being "worked", and is high enough that you might actually be able to finish a few before those LMHers working in the 900M - 1B ranges finish there and move down.

Whatever you decide to do, welcome to GIMPS!!!

[Uncwilly]

Originally Posted by chalsall

But if you must (and for your other slower machines) may I suggest you work in the 889M range?

That is the range that I meant to refer to. The 62 to 63 was an example. 63 to 64 will take 2 times as long.

Operation Billion digit is likely to open up a range for older CPU's.

And for Rodrigo's benefit, the smaller the number you are working on, the longer it will take to a given bit level.

[Rodrigo]

Originally Posted by chalsall

But keep in mind there are *no* exponents at below 63 "bits" other than very low exponents.

Rodrigo, I would really encourage you to not even bother bringing your P75 online. But if you must (and for your other slower machines) may I suggest you work in the 889M range? It's not currently being "worked", and is high enough that you might actually be able to finish a few before those LMHers working in the 900M - 1B ranges finish there and move down.

Whatever you decide to do, welcome to GIMPS!!!

chalsall,

Thanks very much for the welcome!

I checked out the table you linked to. Trying to get a handle on it. The numbers down the columns -- do they represent the amount of work that remains to be done in that range, or the amount of work that's already been done?

A bit confused here because I see that (using your example) there's nothing at 62 bits in the 889M range, which suggests that all the work there IS finished. But then there's nothing again at 66 bits, which suggests that no work HAS been done there.

Rodrigo

[Rodrigo]

Originally Posted by Uncwilly

And for Rodrigo's benefit, the smaller the number you are working on, the longer it will take to a given bit level.

Uncwilly,

Did I hear you right that it takes longer to work a smaller number?

Rodrigo

[Rodrigo]

Uncwilly,

Thanks for all the answers -- you've cleared up a bunch of questions for me -- and for pointing me in various useful directions. I'll explore the LMH section.

Sorry for yet another elementary question, but would you be so kind as to explain to a newbie why it's more valuable to work in the 889M or 930M range, as suggested in this thread, than in (let's say) the 834M range, which if I'm reading this table http://www.mersenne.info/?s=830000000&d=2&t=1 right, still has plenty of work left to do at 63-64?

Rodrigo

[Uncwilly]

Originally Posted by Rodrigo

Did I hear you right that it takes longer to work a smaller number?

Time for a little math.

One of the special properties of any Mersenne number is, that if it has a factor, the factor must be in the form 2kp+1. P is the exponent of the mersenne number: 2^p-1 To test a smaller number, like 971 to check for factors that are say 64 bits long, means that the value for k get really big. This means that there are more numbers to test, like upto 9,498,838,348,975,052.
Whereas if the exponent was for a number in the prize winning range, k might only equal 27,763,586,862
That is a lot fewer numbers to test.

(The numbers posted are based upon quickie calculation. YMMV)

[Uncwilly]

Originally Posted by Rodrigo

Sorry for yet another elementary question, but would you be so kind as to explain to a newbie why it's more valuable to work in the 889M or 930M range, as suggested in this thread, than in (let's say) the 834M range, which if I'm reading this table http://www.mersenne.info/?s=830000000&d=2&t=1 right, still has plenty of work left to do at 63-64?

I personally don't keep upto date on what is currently 'reserved' up there. Others like NBtarheel might. Over in the LMH forum, someone may have said that they will working there.

[chalsall]

Originally Posted by Rodrigo

I checked out the table you linked to. Trying to get a handle on it. The numbers down the columns -- do they represent the amount of work that remains to be done in that range, or the amount of work that's already been done?

The latter. Each column shows how many exponents are *at* that "bit level". So, for the 889M range, you'll see that there are 23,623 exponents which have been trial factored to 63 "bits", 65 to 64 bits, 17 to 65 bits, 1 to 65 bits and 1 beyond 70 bits.

Originally Posted by Rodrigo

A bit confused here because I see that (using your example) there's nothing at 62 bits in the 889M range, which suggests that all the work there IS finished. But then there's nothing again at 66 bits, which suggests that no work HAS been done there.

Hopefully my above explains it, but to be explicit, what this means is none of the exponents have been trial factored to 66 bits (although two have been TFed past 66 bits). The data is a summery of, for example, Factoring Effort report at Mersenne.org (AKA PrimeNet Server).

Originally Posted by Rodrigo

Sorry for yet another elementary question, but would you be so kind as to explain to a newbie why it's more valuable to work in the 889M or 930M range, as suggested in this thread, than in (let's say) the 834M range, which if I'm reading this table http://www.mersenne.info/?s=830000000&d=2&t=1 right, still has plenty of work left to do at 63-64?

Absolutely nothing. I simply suggested that range because it's the highest one in the 800M to 900M range, and the bigger the exponent the faster it will complete. And I know that there is absolutely no work currently being done in the 800M to 900M range, so it's safe for you to "get your feet wet" there.

However, it would be advisable for you to "claim a range" over on the LMH Forum, so those working down above you don't step on you (although there's nothing to guarantee that won't happen, but most LMH try to be polite to each other). Please try to not take too large a range though, since you'll only be doing a few a week, while others do quite a bit more. I'm currently producing 50,000 a day in the 600M and 700M ranges, for example.