My Apologies
 

Hello,

About a month ago I had posted here seeking information on whether it was possible to set up multiple computers to help each other out. Everybody who answered my rookie questions ([URL]http://mersenneforum.org/showthread.php?t=13515[/URL]) was most helpful and patient.

I want to thank all who took the time to help me figure things out. But more importantly, I want to offer my sincere apologies for leaving everything hanging for these several weeks as we dealt with certain private matters.

But now I am back, hopefully for good, and I'm ready to start my PCs contributing to GIMPS.

I welcome your guidance on selecting work and setting it up. Here is my roster of available machiness:

Pentium 75MHz, 128MB RAM tower, Windows for Workgroups 3.11
Pentium 233MHz, 80MB RAM notebook, Windows 98 Second Edition
Pentium II MMX 400MHz, 384MB RAM tower, Windows 98 Standard Edition
Pentium Dual-Core CPU T4200 @ 2.00GHz, 4GB RAM laptop, Vista 64-bit
Pentium Dual CPU E2200 @2.20GHz, 4GB RAM tower, Vista 64-bit

If I recall correctly, NBtarheel_33 suggested that for the P75 I select trial factoring in the following ranges: 932-933M, 934-935M, 936-937M, 938-939M.

I intend to go with the recommendations for the various PCs that were offered in that thread. Please let me know if circumstances have changed in the interim, and what you'd now recommend that I have each computer do for GIMPS. I'm open to all suggestions. My only preference is that I do want the two Vistas to seek prizes. :smile: I'll have to look this up as I get reacquainted with the details, but I think there were some smaller interim awards as well, and they're O.K. too.

Thank you very much.

Rodrigo

[QUOTE=Rodrigo;222909]I
I welcome your guidance on selecting work and setting it up. Here is my roster of available machiness:

Pentium 75MHz, 128MB RAM tower, Windows for Workgroups 3.11
Pentium 233MHz, 80MB RAM notebook, Windows 98 Second Edition
Pentium II MMX 400MHz, 384MB RAM tower, Windows 98 Standard Edition
Pentium Dual-Core CPU T4200 @ 2.00GHz, 4GB RAM laptop, Vista 64-bit
Pentium Dual CPU E2200 @2.20GHz, 4GB RAM tower, Vista 64-bit
[/QUOTE]

Your slower 3 machines have a combined throughput of 1-2% of the other two machines. Consider retiring them (at, the very least, the slowest two).

For the top two, select first time LL. Current state-of-the-art CPUs are underpowered for prize-winning searches.

[QUOTE=Rodrigo;222909]I welcome your guidance on selecting work and setting it up. Here is my roster of available machiness:

Pentium 75MHz, 128MB RAM tower, Windows for Workgroups 3.11
Pentium 233MHz, 80MB RAM notebook, Windows 98 Second Edition
Pentium II MMX 400MHz, 384MB RAM tower, Windows 98 Standard Edition
Pentium Dual-Core CPU T4200 @ 2.00GHz, 4GB RAM laptop, Vista 64-bit
Pentium Dual CPU E2200 @2.20GHz, 4GB RAM tower, Vista 64-bit

Rodrigo[/QUOTE]

My opinion:
The first 3 really are more or less only capable of low level TF factoring (< 64 bits). That being said this project is weeks away from completing all TF below 64 bits. Any assignment other than TF would either take VERY long or you would not have enough RAM. If you still want to run them you should still select TF-LMH anyway.
- Select TF-LMH for the first 3.

The last 2 are much newer and more powerful and truly could handle any work type quite well. The current biggest need is Double Checking.
- So I would suggest DC (Double Checking) for both cores of the last 2. But if you would rather run LL or PM1 or ECM they would handle those as well. NOTE: If you choose PM1 or ECM you should allocate at least 600M of RAM for each core with those work types.

[quote=Rodrigo;222909]I welcome your guidance on selecting work and setting it up.[/quote]Rule #1: Enjoy what you do. Pick types of work that please you. (Perhaps sample at least one assignment of each type before deciding your long-term preferences.)

Rule #1 may conflict with some other advice you get, but if you faithfully follow it, you'll be happier in the long run.

GIMPS throughput will take care of itself as long as you follow Rule #1. Throughput's not a deity that demands devotion from everyone. However, it does have some [I]very[/I] devoted and aggressive devotees, so keep an eye on the predicted completion dates for your assignments. The farther in the future the predicted date, the more likely you are to be hassled about it by throughput-devotees.

[QUOTE=Rodrigo;222909]My only preference is that I do want the two Vistas to seek prizes. :smile:[/QUOTE]I agree with axn : let your 3 old machines rest in peace.

Since you want to "seek prizes" you have only two options for the two Vista machines :
- "First time tests" and
- "World record sized numbers to test".

If you are very lucky and do find one you might share in the 150 000 USD prize that is offered for the first 100 000 000 digit prime (which unless somebody is very lucky or unlikely clever is tens of years off.)

You could also directly try to test 100 000 000 digit primes, each test would take you a few years.

Jacob

[QUOTE=axn;222921]Your slower 3 machines have a combined throughput of 1-2% of the other two machines. Consider retiring them (at, the very least, the slowest two).[/QUOTE]

I *strongly* agree with this. The available exponents with <64 "bits" of TF work to be done will be completed (or, at least, assigned) within two weeks.

Do as cheesehead suggested on another thread, and donate the three slowest machines (which you haven't used for at least a while based on what you've said previously) to a charity.

If you *must* bring these slow machines online, simply select "TF-LMH" on the client during Prime95 installation, and let the server assign work to you. It is not worth your time learning how to "self-assign" when all <64 bit TF work above 200M will be completed in about a week.

(For example, the 932-933M, 934-935M, 936-937M, 938-939M ranges are already long gone....)

[quote=axn;222921]For the top two, select first time LL. Current state-of-the-art CPUs are underpowered for prize-winning searches.[/quote]

axn,

Thank you for the idea, I appreciate it.

But I'm not sure what you mean that (even) state-of-the-art CPUs are underpowered for prize-winning searches. Maybe I misunderstood something. In your opinion, will a realistic search have to wait for even more powerful CPUs than, say, the Intel Core i9 ? That would rate a "Wow!"

Rodrigo

[quote=petrw1;222925]- Select TF-LMH for the first 3.

The last 2 are much newer and more powerful and truly could handle any work type quite well. The current biggest need is Double Checking.
- So I would suggest DC (Double Checking) for both cores of the last 2. But if you would rather run LL or PM1 or ECM they would handle those as well. NOTE: If you choose PM1 or ECM you should allocate at least 600M of RAM for each core with those work types.[/quote]

petrw1,

Thanks very much for the tip about the RAM allocation. This sort of suggestion with details is just what I'm looking for!

Rodrigo

[QUOTE=Rodrigo;222995]axn,

Thank you for the idea, I appreciate it.

But I'm not sure what you mean that (even) state-of-the-art CPUs are underpowered for prize-winning searches. Maybe I misunderstood something. In your opinion, will a realistic search have to wait for even more powerful CPUs than, say, the Intel Core i9 ? That would rate a "Wow!"

Rodrigo[/QUOTE]
Wow, indeed. Currently, the lowest open prize is for 100million (exponent>333M) digit prime. The fastest quad-cores will take nearly a year to complete a single test, running on all four cores!!! (see this thread for benchmarks: [url]http://www.mersenneforum.org/showthread.php?t=13185[/url])

[quote=cheesehead;222936]Rule #1: Enjoy what you do. Pick types of work that please you. (Perhaps sample at least one assignment of each type before deciding your long-term preferences.)[/quote]

cheesehead,

Your philosophy is very attractive. Why get into something like this if it's going to become an ordeal.

And thanks for the heads-up on other folks' preferences. I see now that there is more than one way to approach this project. Sure don't want to frustrate anybody -- I'll try to be as unobtrusive as possible, while still having the older PCs contribute in their limited, 1998 :mike: sort of way.

Rodrigo

[quote=S485122;222938]
Since you want to "seek prizes" you have only two options for the two Vista machines :
- "First time tests" and
- "World record sized numbers to test".
Jacob[/quote]

Jacob,

Merci bien!

Does the Prime95 software allow you to select "world record sized numbers" as the assignment?

Rodrigo

[quote=chalsall;222990]If you *must* bring these slow machines online, simply select "TF-LMH" on the client during Prime95 installation, and let the server assign work to you. It is not worth your time learning how to "self-assign" when all <64 bit TF work above 200M will be completed in about a week.

(For example, the 932-933M, 934-935M, 936-937M, 938-939M ranges are already long gone....)[/quote]
chalsall,

Thanks for the tip, and for the update on the available ranges. I'll keep things on the simpler side, at least while I'm getting acquainted with the whole process.

Do I remember it right that it's recommended to go into the LMH section to reserve work? I'm guessing that one would do that in addition to (rather than instead of) having the server assign the work.

Rodrigo

[quote=axn;222997]Wow, indeed. Currently, the lowest open prize is for 100million (exponent>333M) digit prime. The fastest quad-cores will take nearly a year to complete a single test, running on all four cores!!! (see this thread for benchmarks: [URL]http://www.mersenneforum.org/showthread.php?t=13185[/URL])[/quote]

axn,

I'll check out that link right away, thank you!

Rodrigo <trying to learn fast>

[quote=Rodrigo;223001]chalsall,

Thanks for the tip, and for the update on the available ranges. I'll keep things on the simpler side, at least while I'm getting acquainted with the whole process.

Do I remember it right that it's recommended to go into the LMH section to reserve work? I'm guessing that one would do that in addition to (rather than instead of) having the server assign the work.

Rodrigo[/quote]
If you have PrimeNet automatically assign you LMH work, then there's no need to reserve the work in the forum; that's all handled automatically by the server. The only times you need to reserve it in the forum is when you want to take a large range that you selected yourself (rather than the server selecting the exponents for you) and thus need to make sure nobody else doing the same thing "steps on" your range. The independent LMH workers stay clear of the server's work areas on their own, so anything you get from it is all clear.

[quote=mdettweiler;223004]If you have PrimeNet automatically assign you LMH work, then there's no need to reserve the work in the forum; that's all handled automatically by the server. The only times you need to reserve it in the forum is when you want to take a large range that you selected yourself (rather than the server selecting the exponents for you) and thus need to make sure nobody else doing the same thing "steps on" your range. The independent LMH workers stay clear of the server's work areas on their own, so anything you get from it is all clear.[/quote]

mdettweiler,

Ah, that's very good to know -- thanks!

I've learned so much today; the fog is slowly lifting.

Rodrigo

[QUOTE=Rodrigo;222909]Hello,

About a month ago I had posted here seeking information on whether it was possible to set up multiple computers to help each other out. Everybody who answered my rookie questions ([URL]http://mersenneforum.org/showthread.php?t=13515[/URL]) was most helpful and patient.

I want to thank all who took the time to help me figure things out. But more importantly, I want to offer my sincere apologies for leaving everything hanging for these several weeks as we dealt with certain private matters.

But now I am back, hopefully for good, and I'm ready to start my PCs contributing to GIMPS.


Rodrigo[/QUOTE]

Welcome back, Rodrigo! The advice given above is very sound. I'm did not read all of the posts, but to answer your question posted above- it is possible to select "world-record size first time tests" from the work type menu. These exponents will not qualify for a prize 'until' a 100M digit has been found, then money will be allocated out for all the primes found after the allocation of the previous prize. However, it is important to note two things about this caveat. Firstly, the amount given will likely be small. Second, this could take a long, long time as a 100M digit candidate takes years to test on the average computer. I myself have two computers working for GIMPS; the faster (quad core) is doing first-time tests of world record size while the second will be moved to double checks when it finishes its current assignments. Hope this helps!

Side note: Beautiful, glorious, correct grammar!

In case you find out that your older machines are too slow to get into GIMPS, please consider also donating your CPU cycles to other projects, like (hint! hint!) Operation Billion Digits... :smile: We just opened some new ranges at 62 bits (and factoring billion digits Mersenne numbers is faster than 100 million digits).

Luigi

[QUOTE=ET_;223034]In case you find out that your older machines are too slow to get into GIMPS, please consider also donating your CPU cycles to other projects, like (hint! hint!) Operation Billion Digits... :smile: We just opened some new ranges at 62 bits (and factoring billion digits Mersenne numbers is faster than 100 million digits).

Luigi[/QUOTE]

I agree. I think that the big guns shouldn't do OBD because it is too low-priority. They should do DC instead: that is what is needed ATM.

Primeinator,

Well, thanks for the comments on my grammar! I imagine that most of the folks here are excellent at math. For me, everything up to Probability was interesting and fun, but I knew I wasn't cut out for sophisticated mathematics when the time came to do trigonometry, and then calculus. Doing integrals was enjoyable, but derivatives were just incomprehensible to me (or was it the other way around -- been a long time). And when homework started to consist of determining the volume of cones :yawn:, I realized that my career was going to involve working with words and not numbers. :smile:

Thank you, too, for explaining the world-record tests, I appreciate it. Are those related to the numbers that qualify for the smaller Mersenne Prime Awards ([URL="http://www.mersenne.org/legal/#awards"][COLOR=#800080]http://www.mersenne.org/legal/#awards[/COLOR][/URL])?

One more question, if that's O.K.--

About those smaller, interim awards for finding sub-100M primes: When getting work for your (modern) PCs, would you do anything differently if you were pursuing a Mersenne Prime Award vs. the 100 Million Digit Prime Award? The reason I ask is that I'm thinking that maybe a more modest goal would be in order, but maybe these two functions just go together naturally and there's no need to choose one or the other.

Gratefully,

Rodrigo

[quote=ET_;223034]In case you find out that your older machines are too slow to get into GIMPS, please consider also donating your CPU cycles to other projects, like (hint! hint!) Operation Billion Digits... :smile: We just opened some new ranges at 62 bits (and factoring billion digits Mersenne numbers is faster than 100 million digits).

Luigi[/quote]
Luigi,

Thank you for the suggestion. OBD is definitely in the running. Even if the ultimate goal is that much further away, knowing that my older machines are completing more tasks in a given time is an attractive option.

How would I get OBD work assigned to my PCs?

Rodrigo

[quote=Rodrigo;223194]About those smaller, interim awards for finding sub-100M primes: When getting work for your (modern) PCs, would you do anything differently if you were pursuing a Mersenne Prime Award vs. the 100 Million Digit Prime Award? The reason I ask is that I'm thinking that maybe a more modest goal would be in order, but maybe these two functions just go together naturally and there's no need to choose one or the other.[/quote]
The Mersenne Prime Award ($3000) is awarded for the discovery of *any* new Mersenne prime with less than 100 million digits. The "First Time Tests" and "World Record Tests" work options both assign tests within this range. In many cases both would assign the same numbers, though "Workd Record Tests" is specifically limited to numbers above the largest known prime, whereas "First Time Tests" can give out smaller "fill-in" work as available that hasn't yet been completed. Either could just as easily produce a prime eligible for the Mersenne Prime Award. Additionally, the LL-10M choice, which produces first time tests of 10 million digits or greater (left over from the last big award for 10 million digits), nowadays is almost exactly the same as "First Time Tests" because almost all numbers less than 10 million digits have already been tested once. All three options produce numbers that will test in about 3-4 weeks on one core of a modern CPU.

The 100 Million Digit Prime Award ($50000) is awarded for the discovery of the *first* prime of 100 million digits or greater. This means numbers much, much bigger than the current world record and the leading edge of "First Time Tests" progression. You must select the LL-100M choice to get numbers of this size, though they're so big that they'll take over a year to do even with four cores of a modern quad working on the same test. (Only a few weeks ago was the very first of these actually completed.)

So the two goals are in fact mutually exclusive--100 million digit numbers are specifically excluded from the $3000 Mersenne Prime Award because they've already got the $50000 award reserved for them. Note that the 100 million digit numbers, due to their size being so far ahead of today's computer processing capability, have an extremely remote chance of finding a prime any time soon. Even the "normal" first-time LL tests have a rather remote chance of any individual user finding a prime, though at least the chances are enough that GIMPS as a whole has a decent chance of finding one every couple of years.

Even though the potential rewards of working for the 100 million digit prize is much greater, the odds of actually achieving it are infinitesimal. With the Mersenne Prime Award, the dollar amount is comparatively small, but you'd at least have sort of a chance of actually getting it. :smile: It's like playing regular lottery vs. Mega Millions--the chances of winning either is remote, but the latter is so much harder that you'd probably have better luck trying to win the smaller lottery multiple times. At GIMPS, the odds against winning the 100 million digit award (i.e., "mega millions") are even more disparate from the Mersenne Prime Award (i.e. "regular lottery") than they would be with a real lottery.

mdettweiler,

Wonderful explanation from beginning to end, thank you -- just the sort of thing I was hoping for! It was a HUGE help.

Rodrigo

[QUOTE=Rodrigo;223195]Luigi,

Thank you for the suggestion. OBD is definitely in the running. Even if the ultimate goal is that much further away, knowing that my older machines are completing more tasks in a given time is an attractive option.

How would I get OBD work assigned to my PCs?

Rodrigo[/QUOTE]

1 - Download the program suitable for your platform from [URL="http://www.moregimps.it/billion/download1.php"]here[/URL]
1 - Choose one or more exponents from [URL="http://www.moregimps.it/billion/expo_i.php?id=bit_depth"]here[/URL]
2 - Determine how deep you want trial-factor the Mersenne number (notice that taking it from 63 to 64 bits needs twice the time needed to take it from 62 to 63) making some tests.
3 - Check out the number [URL="http://www.mersenneforum.org/showthread.php?p=223042#post223042"]here[/URL]
4 - Report your results [URL="http://www.mersenneforum.org/showthread.php?p=223166#post223166"]here[/URL]

And

5 - Ask me for whatever question about it. :smile:

[edit]
I'm afraid the actual versions on the website are optimized for core processors. Give them a try and let me know, we have probably to recompile the source code on your platform :sad:

Luigi

Luigi, this is great! You've provided the "complete package." :smile:

I have yet to look at the other steps, but I already have a follow-up question on the first one. On the page where you select the version of the program, which one would you recommend for a lowly P75 and a P233? Basically I'd like to know if there is a reason to choose Factor3_1_Win (the slowest version) instead of, for example, Factor4_Win (and would I choose the SSE2 variety).

In case it makes a difference in the selection, the P75 is running DOS 6.20 and Windows for Workgroups 3.11; while the P233 is running Windows 98 Second Edition.

[edit] Oh, wait -- I just remembered you had an edit that addressed the question of how to select the version. Maybe I should just pick one and see what happens?

Thanks!

Rodrigo

[QUOTE=Rodrigo;223195]Thank you for the suggestion. OBD is definitely in the running. Even if the ultimate goal is that much further away, knowing that my older machines are completing more tasks in a given time is an attractive option.[/QUOTE]

"Even if the ultimate goal is that much further away...

Yeah. Hundreds of years...

I'm not trying to be negative here, but I have to ask what is the point in investing time and energy into something which is truly pointless?

By the time a billion digit P exponent (2^P-1) will be tested we will (hopefully) have quantum computers able to do this kind of work in a tiny fraction of the time our current technology can.

So why burn oil having a P75 work on this?

chalsall,

A pertinent question.

I'm considering OBD largely because I get the impression that if I put the P75 on GIMPS there will be grumbling about how it's slowing down the train. Moreover, it sounds like over at OBD it can accomplish particular tasks quicker than in the GIMPS "mainstream."

But why even bother putting the P75 online? Let me backtrack some here.

We might call it a kind of virtuous circle. When I first heard of GIMPS, and before knowing any of the details, I got really excited about the prospect of putting my older PCs back in service doing something useful instead of gathering dust.

That prospect then motivated me (a) to upgrade those PCs with additional memory and network cards, and (b) to set up a home network so that each of the computers can report in when needed. [B]In the process, I've learned a ton about computing.[/B] So the mere fact of hearing about GIMPS has already had positive external effects, which would not have occurred otherwise.

Now I'd like to make use of this newfound knowledge (completing the circle) by contributing them to GIMPS and/or OBD, the very concept that prompted me to acquire that knowledge in the first place. If I now fail to put them on one or the other of these projects, it will be as if all of that effort and learning had been for naught. What economists call the "psychic value" of the activity will disappear.

I hope I've explained adequately where I'm coming from.

Rodrigo

[QUOTE=Rodrigo;223275]Now I'd like to make use of this newfound knowledge (completing the circle) by contributing them to GIMPS and/or OBD, the very concept that prompted me to acquire that knowledge in the first place. If I now fail to put them on one or the other of these projects, it will be as if all of that effort and learning had been for naught. What economists call the "psychic value" of the activity will disappear.

I hope I've explained adequately where I'm coming from.[/QUOTE]

I appreciate your above.

I hope you et al will appreciate this...

1. I am very pleased that you have learnt from this experience.

2. No teacher expects to be compensated directly from their student's learning.

2.1. A teacher will, however, always hope that their students go on to help and teach others.

3. Your (and many others') slow machines would be much better off being donated to a charity which can route them to those who have no computing resources currently.

3.1. No disrespect intended, but telling the truth -- one of my machines can do in a few minutes what one of your older machines would take days to do -- with *much* less power consumed per work unit.

From this, then, may I please advise you (and possibly others) to...

1. Donate your slow machines to a charity which will see that they end up in the hands of those who need them.

2. Use the knowledge you've gained here to support these (and possibly other) computers; where ever they end up.

2.1. You could still help GIMPS in this case. But only slowly; and while the machine was doing more than consuming energy and creating heat. :smile:

The money you save on your electric bill by not running the older machines would quickly add up to the purchase price of a new machine which would outpace the combined output of your old machines by 10's if not 100's of times and would give you much more work done over the long run.

[quote=chalsall;223281]3.1. No disrespect intended, but telling the truth -- one of my machines can do in a few minutes what one of your older machines would take days to do -- with *much* less power consumed per work unit.
[/quote]

chalsall,

Thank you for the advice. Just one question --

Help me to understand this. My machines are a P-75 :showoff:, a P-233, a PII-400, a P-T4200 (2.00GHz), and a P-E2200(2.20GHz). For the sake of simplicity, let's suppose that all together these five PCs can produce 1.00 units of output for GIMPS. Let's say that the first three account for 0.01 of the total, with the last two contributing 0.99. (The actual proportions don't matter; they could be 0.0001 and 0.9999 and the issue would remain the same.)

I'm no math whiz, but in terms of helping GIMPS to get its work done, I'm not sure how 0.99 units of contributed output might be preferable to 1.00 units of contributed output. (I do understand about the matter of kW per unit of work, but that's not what I'm asking here.)

I do appreciate your taking the time to explain. :smile:

Rodrigo

As far as the saving money aspect is concerned, doing 0.99 units for a year, and then 1.50 units after upgrading will work out to more total work done in the long run. (My 1-year figure is just a wild guess on my part - and this is all assuming you 1) even care about the electricity cost and 2) would put the savings of not running the old machines towards a new one)

As far as GIMPS is concerned, the more work the better, regardless of the cost. However, if you hold out the 0.01 units with the goal of adding 0.50 units in the future, it's better for GIMPS in the long run. I'm sure there's a name for this idea, but I'm no economist. :)

Of course, if money were no object, you could just keep running the old machines AND buy a new one, and end up with 1.51 units of work being done. ;)

[QUOTE=Rodrigo;223286]I'm no math whiz, but in terms of helping GIMPS to get its work done, I'm not sure how 0.99 units of contributed output might be preferable to 1.00 units of contributed output.[/quote]
It isn't preferable for GIMPS -- but it is preferable for you. Don't think that we're saying that the offer of computation is not wanted. Some computation is better than no computation!

However, with technology that old, the cost of electricity is a very real factor to be considered (again, for you). Eventually, it is up to you to make the decision to see if it is worth it or not to run these old computers. However, we just hope that the decision you make is an informed one.

Consider this: if you overclock your two fastest computers by a paltry 25MHz, you would have already accomplished more computation per second than your slowest 3 computers!

[QUOTE=Rodrigo;223286]I'm no math whiz, but in terms of helping GIMPS to get its work done, I'm not sure how 0.99 units of contributed output might be preferable to 1.00 units of contributed output. (I do understand about the matter of kW per unit of work, but that's not what I'm asking here.)[/QUOTE]

You might consider the global impact you're having contributing to GIMPS.

You are consuming energy, and generating heat, while you inefficiently contribute to GIMPS (or any other distributed computing effort) using these ancient machines.

If you're comfortable with that, then fine.

But please know that that energy could probably go a lot further being used elsewhere....

[QUOTE=chalsall;223273]"Even if the ultimate goal is that much further away...

Yeah. Hundreds of years...

I'm not trying to be negative here, but I have to ask what is the point in investing time and energy into something which is truly pointless?

By the time a billion digit P exponent (2^P-1) will be tested we will (hopefully) have quantum computers able to do this kind of work in a tiny fraction of the time our current technology can.

So why burn oil having a P75 work on this?[/QUOTE]

What about algorithm enhancement? :smile:

Thank to OBD, we discovered lots of low-hanging fruits (factors) of billion digits Mersenne numbers. Thank to Oliver (TheJudger) we now have a Mersenne factoring program that can reach 95 bit factors. But reaching 95 bits depth even on OBD numbers would be a tough task. Luckily enough, our previous nonsensical work provided him with factors that, multiplied together, could simulate a 90-95 bit huge factor to test his program.

Nothing of the previous paragraph would have happened without slow computers thrown into OBD...

Luigi

[QUOTE=Rodrigo;223272]
In case it makes a difference in the selection, the P75 is running DOS 6.20 and Windows for Workgroups 3.11; while the P233 is running Windows 98 Second Edition.

Rodrigo[/QUOTE]

I'm afraid you should then abandon the P75, or install some light Linux distro on it: Factor_x makes use of GMP (Gnu MultiPrecision) Library.

GMP is a set of math functions written in C and optimized in Assembler: it is distributed as source code, so that the end user can configure each call using the best enhancements available for his/her architecture.
The bad news: GMP only compiles under Linux, Unix or Cygwin/MinGW (the last being a 32/64 bit framework to be installed under Windows from 98 up).

So it's up to you to choose between the retirement of your P75 or a new Linux hope... :smile:

Luigi

[quote=ET_;223324]I'm afraid you should then abandon the P75, or install some light Linux distro on it: Factor_x makes use of GMP (Gnu MultiPrecision) Library.

GMP is a set of math functions written in C and optimized in Assembler: it is distributed as source code, so that the end user can configure each call using the best enhancements available for his/her architecture.
The bad news: GMP only compiles under Linux, Unix or Cygwin/MinGW (the last being a 32/64 bit framework to be installed under Windows from 98 up).

So it's up to you to choose between the retirement of your P75 or a new Linux hope... :smile:

Luigi[/quote]I first ran Linux on a 386 and it's likely that I still have the installation media for distributions which work well on machines of that
age. In fact, my PPro-233 (which predates a P75) still works and runs an old version of RedHat though it's not been powered up for almost a year.

If you (chalsall) want to install Linux on your P75 and want an old version better suited for elderly machines please contact me.

Paul

[QUOTE=ET_;223324]I'm afraid you should then abandon the P75, or install some light Linux distro on it: Factor_x makes use of GMP (Gnu MultiPrecision) Library.

GMP is a set of math functions written in C and optimized in Assembler: it is distributed as source code, so that the end user can configure each call using the best enhancements available for his/her architecture.
The bad news: GMP only compiles under Linux, Unix or Cygwin/MinGW (the last being a 32/64 bit framework to be installed under Windows from 98 up).

So it's up to you to choose between the retirement of your P75 or a new Linux hope... :smile:

Luigi[/QUOTE]
Luigi,

That's a bummer.

I'll have to investigate whether there is a way to multiboot an ancient machine such as that P75. I just installed a second hard disk on it. If multibooting is possible, I could move all the data files over to the new HDD, defragment the old HDD, and then create a new partition for Linux. What size partition would be needed by the version of Linux that you mention in order to function, and to run Factor_x ?

Rodrigo

P.S. To post this, I had to log in as unregistered and changing my name slightly because every time I logged in to the forum and then tried to reply to your post, I got logged out.

[QUOTE=xilman;223326]I first ran Linux on a 386 and it's likely that I still have the installation media for distributions which work well on machines of that
age. In fact, my PPro-233 (which predates a P75) still works and runs an old version of RedHat though it's not been powered up for almost a year.

If you (chalsall) want to install Linux on your P75 and want an old version better suited for elderly machines please contact me.

Paul[/QUOTE]
Paul,

It was me, Rodrigo, and not chalsall, who might need that Linux. :-)

Assuming that I can multiboot on the P75, what size partition would I need to run Linux and then Luigi's program for OBD?

Rodrigo

Luigi,

One other quick question:

Is there any other software that will enable that P75 to participate in OBD without having to install a new operating system?

Grazie!

Rodrigo

[quote=Rodrigo2;223358]Paul,

It was me, Rodrigo, and not chalsall, who might need that Linux. :-)

Assuming that I can multiboot on the P75, what size partition would I need to run Linux and then Luigi's program for OBD?

Rodrigo[/quote]When I first ran Linux on a 386 it dual-booted Win 3.1 and everything fitted on a 120MB disk, though it was rather cramped. A 1995-vintage Linux will fit into a few tens of megabytes but I would recommend not running X11 (i.e. a window system) but keep it entirely text mode. If you are familiar with DOS you'll soon learn how to drive a Linux system in text mode.

I've no idea how much space Luigi's program requires. He will have to answer that one.


Paul

P.S. I may still have a P75 hidden away in my loft. I certainly still have the 386 system, though the processor was upgraded with a 486sx before the machine went into storage.

[QUOTE=xilman;223369]When I first ran Linux on a 386 it dual-booted Win 3.1 and everything fitted on a 120MB disk, though it was rather cramped.[/QUOTE]There was one distro that I tried on a 486 that one could launch from DOS. Also, Zipham could boot off a 1.4 floppy. Knoppix and others have live CD boot options.

[QUOTE=ET_;223323]What about algorithm enhancement? :smile:[/QUOTE]

What about it? :smile:

[QUOTE=ET_;223323]Thank to OBD, we discovered lots of low-hanging fruits (factors) of billion digits Mersenne numbers.[/QUOTE]

But none that couldn't have been found by faster machines with much less energy being consumed doing so.

[QUOTE=ET_;223323]Thank to Oliver (TheJudger) we now have a Mersenne factoring program that can reach 95 bit factors. But reaching 95 bits depth even on OBD numbers would be a tough task. Luckily enough, our previous nonsensical work provided him with factors that, multiplied together, could simulate a 90-95 bit huge factor to test his program.[/QUOTE]

I agree with this. I have been awed by Oliver's work. It will be the reason my next personal workstation will be a desktop with a high-end NVidia card.

[QUOTE=ET_;223323]Nothing of the previous paragraph would have happened without slow computers thrown into OBD...[/QUOTE]

I *don't* agree with this.

If there was a solid reason for empirical data to be made available, I know at least one person who would have made modern computing resources available to generate that data efficiently....

[QUOTE=Rodrigo2;223358]Assuming that I can multiboot on the P75, what size partition would I need to run Linux and then Luigi's program for OBD?[/QUOTE]

Just wondering...

Why do you need to multi-boot a machine which you're going to dedicate to GIMPS and/or OBD? Wipe the hard-drive, and install Linux as the sole operating system. (It's another learning opportunity.)

But to answer your question, I very recently installed CentOS 5.3 on a machine, and the base (command-line only) install took just under 2 GB of file-space.

A corollary -- the others above are correct; it is possible to boot into Linux using less than a single floppy disc's worth of file-space. If this will be enough to run GIMPS and/or OBD is another question I cannot answer....

[quote=Rodrigo2;223357]every time I logged in to the forum and then tried to reply to your post, I got logged out.[/quote]That was probably because you weren't allowing cookies from mersenneforum.org.

[quote=chalsall;223384]Just wondering...

Why do you need to multi-boot a machine which you're going to dedicate to GIMPS and/or OBD? Wipe the hard-drive, and install Linux as the sole operating system. (It's another learning opportunity.)
[/quote]

chalsall,

Another fair question.

That P75 was my main office computer from 1994 to 2002. Wiping the hard drive would mean losing all the software that came pre-installed from Dell, as there were no floppies or Setup files from which to reinstall. I don't want to lose access to that.

Moreover, all my e-mail from that period is in a proprietary format for software that won't run on any of my other PCs, and I do not want to lose access to it. (Yes, the need to access it does still arise once in a while.)

Finally, and from a psychological viewpoint, I have a chain of PCs dating back to my 1984 Sanyo MBC-555, every one of them still fully functional and in the same state (OS) in which I used them back in their time. It's a pleasure to be able to go back to any of my PCs whenever I wish and recreate the experience of the time. It's a connection to the past. Wiping out WfW3.11 on my P75 would break the chain, and that connection.

Thanks for the scoop on being able to boot into Linux. We'll await ET/Luigi's answer re: OBD.

Rodrigo

[QUOTE=Rodrigo2;223357]Luigi,

That's a bummer.

I'll have to investigate whether there is a way to multiboot an ancient machine such as that P75. I just installed a second hard disk on it. If multibooting is possible, I could move all the data files over to the new HDD, defragment the old HDD, and then create a new partition for Linux. What size partition would be needed by the version of Linux that you mention in order to function, and to run Factor_x ?

Rodrigo

P.S. To post this, I had to log in as unregistered and changing my name slightly because every time I logged in to the forum and then tried to reply to your post, I got logged out.[/QUOTE]

On another thought, I'm afraid that Factor is not well suited for a Pentium machine:
- Version 1 used the Giantint all-C library from Richard Crandall, and was rather slow.
- Versions 2 and 3 assumed a huge use of RAM (one structure was 64 MB in size).
- Version 4 used structures of 256KB to best fit in cache, but was not optimized.
- Version 5 executables are TOO optimized for modern processors, and uses the same structures.

So, if you want to use Factor5, you need at least 4 MB of RAM installed. The program footprint is about 140 KB if statically compiled, or 33 KB with dynamic libraries.

If you can install an old version of Linux for i586 machines, you are still at the beginning: you should then:
- Download the GMP library
- Configure the environment with the approriate script.
- Make, make install and make check the GMP (that is, compile it on your machine).
- Download factor5 sources.
- Recompile Factor5 sources.
- Cross your fingers :smile:

I'm sorry I don't have a Pentium machine to avoid all this hassle and let you use an already compiled version of Factor5.

Luigi

[QUOTE=Rodrigo;223397]Wiping the hard drive would mean losing all the software that came pre-installed from Dell, as there were no floppies or Setup files from which to reinstall. I don't want to lose access to that.[/QUOTE]

Get over it. You never had ownership of that software. You simply had a license.

[QUOTE=Rodrigo;223397]Moreover, all my e-mail from that period is in a proprietary format for software that won't run on any of my other PCs, and I do not want to lose access to it. (Yes, the need to access it does still arise once in a while.)[/QUOTE]

Then you should understand that every time you access that system there is a chance it will crash -- and you should try to transfer the data from it to a reliable file system as soon as you can.

[QUOTE=Rodrigo;223397]Finally, and from a psychological viewpoint, I have a chain of PCs dating back to my 1984 Sanyo MBC-555, every one of them still fully functional and in the same state (OS) in which I used them back in their time. It's a pleasure to be able to go back to any of my PCs whenever I wish and recreate the experience of the time. It's a connection to the past. Wiping out WfW3.11 on my P75 would break the chain, and that connection.[/QUOTE]

I resonate with what you say.

I have deep emotional connections with machines going back more than 30 years (my first machine was a TRS-80 Model 1 (Z80 processor)).

But, importantly, I don't have a need to have my old machines around me currently because I've long since backed up the data produced by them...

I can visit my old friends in museums like [URL="http://www.bletchleypark.org.uk/"]Bletchley Park[/URL] knowing that they have done their job well...

[quote=enderak;223287]As far as the saving money aspect is concerned, doing 0.99 units for a year, and then 1.50 units after upgrading will work out to more total work done in the long run. (My 1-year figure is just a wild guess on my part - and this is all assuming you 1) even care about the electricity cost and 2) would put the savings of not running the old machines towards a new one)
[/quote]
enderak,

You're right -- the price of running these old machines is not a big consideration in my case. As it is, two of them are running about half the time anyway for other reasons, so the additional cost of running them 24/7 is not as great as one might think.

[quote=enderak;223287]As far as GIMPS is concerned, the more work the better, regardless of the cost. However, if you hold out the 0.01 units with the goal of adding 0.50 units in the future, it's better for GIMPS in the long run. I'm sure there's a name for this idea, but I'm no economist. :)
[/quote]

The issue that you describe does have a name in economics -- "future value" vs. "present discounted value." :smile:

[quote=enderak;223287]
Of course, if money were no object, you could just keep running the old machines AND buy a new one, and end up with 1.51 units of work being done. ;)[/quote]

Getting a new machine is extremely tempting, and we may well get to that point... although, it wouldn't be because of the savings gained from not running the old PCs. :wink: I would do it (well, at least in part, anyway) to contribute a new multicore machine to GIMPS. Maybe raising that 1.51 figure to 3.51 or somewhere around there...

Rodrigo

[quote=cheesehead;223395]That was probably because you weren't allowing cookies from mersenneforum.org.[/quote]
cheesehead,

Hmmm -- I wasn't doing anything different from any other time. Hadn't changed any settings from the night before. And now again things are fine. <scratching head>

Anyway, as I've said previously, I like the approach you take to the project, but I wouldn't want to become an obstacle for the speed enthusiasts. What would be an out-of-the way range where my senior citizen PC could work for a while, where nobody's likely to come barreling through soon and trip over it?

Thanks!

Rodrigo

[quote=ET_;223403]On another thought, I'm afraid that Factor is not well suited for a Pentium machine:
[...]
I'm sorry I don't have a Pentium machine to avoid all this hassle and let you use an already compiled version of Factor5.
[/quote]
Luigi,

Thanks for getting back to me on this. If not Pentiums, then which processors are your programs optimized for -- later (or earlier) Intel CPUs, or a different brand altogether?

And, is there any other software out there, from anybody else, that would enable this PC to do OBD?

My P75 has 128MB of RAM in it, FWIW.

Rodrigo

[quote=chalsall;223407]Get over it. You never had ownership of that software. You simply had a license.
[/quote]
chalsall,

That's O.K. I still have the USE of that software, but I would lose that if I were to wipe the drive.

[quote=chalsall;223407]Then you should understand that every time you access that system there is a chance it will crash -- and you should try to transfer the data from it to a reliable file system as soon as you can.
[/quote]
Considering that my e-mail is in an oddball format that no other software can untangle, I am indeed resigned to that possibility. What can I say -- I've learned a lot more about computing since 1994!

I suppose I could print out hundreds of pages, but that hardly seems practical. I could also copy the e-mail directories onto floppies, but it would take hundreds of them -- an administrative nightmare, and then there would still be the question of what other PC could read them.

[quote=chalsall;223407]But, importantly, I don't have a need to have my old machines around me currently because I've long since backed up the data produced by them...

I can visit my old friends in museums like [URL="http://www.bletchleypark.org.uk/"][COLOR=#800080]Bletchley Park[/COLOR][/URL] knowing that they have done their job well...[/quote]

That's a nice feeling. :smile:

But let me illustrate my angle on this: Recently my father handed me a quad-density 5.25" floppy that he had just discovered sitting in a box, asking me to find out what was on it. I put the 800K floppy in my MBC-555 and determined that they were family documents and private correspondence that, in light of events many years later, turned out to be quite significant. He decided that he wanted the files, but because (like a lot of people) he had long since donated his old computers, he didn't have any that could read this diskette.

So I put a 360K diskette in the other drive of my old Sanyo and copied the files over because none of my other PCs can read a quad-density.

Next I ported the 360K 5.25" floppies over to my P75 because it has both a 3.5" A: drive and a 5.25" B: drive; and I copied the files from the 5.25" to the 3.5". But my dad doesn't even have a 3.5" drive anywhere in sight.

A home network had not even been conceived of at that point, so the next step was to take the 3.5" minifloppy with the old files to my Windows 98 tower, where I converted them from WordStar (remember it?) to Microsoft Word format. Finally, I copied the files to a USB flash drive and offered my dad the option of getting home delivery of the flash drive, or to receive the files by e-mail. He chose home delivery. :smile:

This project involved three computers that many folks would consider utterly obsolete and useless; and yet under the circumstances they helped me to accomplish an unexpectedly valuable family research purpose. I doubt that Bletchley Park (or an equivalent U.S. institution) would have allowed me free use of their equipment to get this done, never mind the time and expense involved in getting there. Nostalgia aside, I am glad that I kept the old machines around and in working order.

Rodrigo

[QUOTE=Rodrigo;223412]Considering that my e-mail is in an oddball format that no other software can untangle, I am indeed resigned to that possibility. What can I say -- I've learned a lot more about computing since 1994!

I suppose I could print out hundreds of pages, but that hardly seems practical. I could also copy the e-mail directories onto floppies, but it would take hundreds of them -- an administrative nightmare, and then there would still be the question of what other PC could read them.[/QUOTE]What program, there may be a converter available.

[QUOTE=Rodrigo;223412]But let me illustrate my angle on this: <snip>[/QUOTE]
All the more reason, not to mess with them, no? Keep using them as normal, but leave the number-crunching to newer ones.

[quote=Uncwilly;223414]What program, there may be a converter available.[/quote]
Uncwilly,

It's WinCIM 1.4, the CompuServe e-mail software.

I found converters that would go as far back as WinCIM 2.x, but not 1.4, and in any case they offered to convert either to Outlook Express (which to me was virtually unfathomable) or to Outlook 98 (which I never had).

Rodrigo

[quote=axn;223415]All the more reason, not to mess with them, no? Keep using them as normal, but leave the number-crunching to newer ones.[/quote]
axn,

Is there a danger of screwing up a machine, or the OS, with the use of Prime95 or a similar program?

I do know about the "torture test" aspect of Prime95 (in fact, that's how I first heard of GIMPS :smile:), but I don't know that it's possible to mess up a PC irretrievably with it.

Rodrigo

[QUOTE=Rodrigo;223420]Is there a danger of screwing up a machine, or the OS, with the use of Prime95 or a similar program?
[/QUOTE]

Not really, but that was not what I was referring to. I was talking about the attempts to install linux.

[quote=axn;223422]Not really, but that was not what I was referring to. I was talking about the attempts to install linux.[/quote]

Oh, I see, thanks.

Yeah, I'm inclined to leave it alone in terms of the operating system. But I'd still like to find some out-of-the way spot where the old machine could settle in and contribute (if ever so modestly) to the project. Trial Factoring in the LMH zone has been suggested, and now that we've learned that OBD appears to require changing the OS the choice is getting clearer. I'll be looking into LMH in the next day or two -- hopefully I can find a place there to park my box where it can chisel away without bothering anybody.

Rodrigo

[QUOTE=Rodrigo;223425]Yeah, I'm inclined to leave it alone in terms of the operating system.[/QUOTE]I might be able to dig up the version of Linux that I tried, it allows all of the Linux files to reside on a DOS partition. You boot to DOS. Then you go to the Linux directory and launch it. It loads itself. All of the Linux files are contained in the Linux directory. It also loads support for the FAT file system. It is on an old laptop that I haven't fired up in 3-5 years. (I may have copied all of the files to an external HDD.

[QUOTE=Rodrigo;223410]Luigi,

Thanks for getting back to me on this. If not Pentiums, then which processors are your programs optimized for -- later (or earlier) Intel CPUs, or a different brand altogether?

And, is there any other software out there, from anybody else, that would enable this PC to do OBD?

My P75 has 128MB of RAM in it, FWIW.

Rodrigo[/QUOTE]

IIRC, there was a program called simply Factor that could run in DOS. Other primality proving and factorization programs are maintained by Will Edgington [URL="http://www.garlic.com/~wedgingt/README.html"]here[/URL]. Some of them should run under a DOS extender.

Pentium 4 should run Factor5 and Factor4. Maybe there was a version of Factor4 that run on Pentium III. Our friend WBLIPP (coordinator of OBD) used an old version of Factor (I guess 3.2) on his PPro 233, but I don't recall which OS he had.

Sadly enough, pfgw is based on Woltman's GWNUM library, and can't handle factorization of huge Mersenne numbers (its form evaluator crashes).

Luigi

Luigi,

Thanks for looking, I appreciate it.

If Uncwilly can find the Linux version that works off a DOS partition, then the P-75 may be back as a possibility for OBD. Otherwise, I'll dedicate the P-233 and/or the PII-400 to OBD, using Factor_x for Windows. Which version do you recommend for these CPUs?

Rodrigo

Uncwilly,

So with this version I would simply boot to DOS as normal, and then instead of continuing on to Windows I would load Linux as if it were a DOS application? That would be very cool.

Before we get into further complications such as dealing with a new OS, though, maybe it's better to exhaust simpler possible alternatives. Is there a range where I could park that P-75 to quietly do Trial Factoring for a while and it wouldn't quickly get in anybody's way?

I'm looking (for example) in the 630M ( [URL]http://www.mersenne.info/?s=630000000&d=2&t=1[/URL] ) and 857M ( [URL]http://www.mersenne.info/?s=850000000&d=2&t=1[/URL] ) ranges. Is a page like this one: [URL]http://www.mersenne.org/report_factoring_effort/?exp_lo=857000000&exp_hi=858000000&bits_lo=0&bits_hi=999&exassigned=1&B1=Get+Data[/URL] sufficient to decide whether I can/should ask to be assigned numbers there?

Rodrigo

I think all of those exponents have been factored past 62 bits (which, if I remember correctly, was the limit of what your oldest machine is capable of).

[QUOTE=chalsall;223273]Yeah. Hundreds of years...[/QUOTE]Or 24.

[quote=http://primes.utm.edu/top20/trends.php]The slope 0.079 (over past 60 years) corresponds to doubling the digits every 3.8 years, or 46 months.[/quote]Horribly out of context, but you can read the rest if you like.

2^d * 12,978,189 = 10^9

Where d is the number of doublings:

2^d = 77.05

d = log_2(77.05)

d = 6.268

Since each doubling takes 46 months, the time to reach 1 billion digit primes is:

T = 6.268 * 46
T = 288 months
T = 24 years

Obviously only a rough estimate, but it's not that long off, lets just call it about 25.

Edit: Oh, and the first 100 million digit prime in 11 to 12 years or so.

If your machines can boot from CD, you might take a look at [URL="http://www.damnsmalllinux.org/"]Damn Small Linux (DSL)[/URL] for a small linux OS that doesn't change anything of the resident OS. You can simply use the HD for storage.

However, heavy CPU usage does take a toll on older machines. Using them 24/7 would not preserve them for future availability.

A last thought: I recently dug a 386/16 out of storage to use for a simple door security setup and discovered that while it was sitting, the CR123 backup battery (soldered to the MB) leaked and destroyed all the surrounding traces, including the internal layers.:sad:

I, too, have some older machines around - I wonder what I could get a dozen TI-99/4A consoles to accomplish. . . :smile:

Take Care,
Ed

[QUOTE=Rodrigo;223501]I'm looking (for example) in the 630M ( [URL]http://www.mersenne.info/?s=630000000&d=2&t=1[/URL] ) and 857M ( [URL]http://www.mersenne.info/?s=850000000&d=2&t=1[/URL] ) ranges. Is a page like this one: [URL]http://www.mersenne.org/report_factoring_effort/?exp_lo=857000000&exp_hi=858000000&bits_lo=0&bits_hi=999&exassigned=1&B1=Get+Data[/URL] sufficient to decide whether I can/should ask to be assigned numbers there?[/QUOTE]

I would suggest you stay away from anything above 200M, as everything still at 63 bits has been officially reserved on the server and/or unofficially on the LMH Forum.

Besides, everything between 310M and 350M will be completed in approximately 48 hours....

[quote=chalsall;223561]I would suggest you stay away from anything above 200M, as everything still at 63 bits has been officially reserved on the server and/or unofficially on the LMH Forum.

Besides, everything between 310M and 350M will be completed in approximately 48 hours....[/quote]
chalsall,

Thanks for the scoop.

How about 190M? [URL]http://www.mersenne.info/?s=100000000&d=1&t=1[/URL]

Which leads me to something I'd been meaning to ask. After all these years and with so many people participating, why is there (apparently) still so much work left undone in the very low ranges? (I.e., [URL]http://www.mersenne.info/?s=0&d=2&t=1[/URL] )

No doubt there is a good reason.

Rodrigo

Ed,

Thank you for the warning re: the backup battery. Bummer. If I understand the issue correctly, this is a problem mainly when the computer has been left unplugged (or is it, simply unused) for long periods of time. Sort of like the batteries in a child's toy, if nobody turns it on for a few months they can leak.

Unfortunately, the P75 came out when CD-ROMs were just getting started, and just like I can't write to it (only read from it), there's no provision in the BIOS for booting from it.

On the TI99/4A's, I can relate. It was a neat time when a whole new world was opening up before us. In addition to my Sanyo MBC-555, I have an Amstrad PC6400DD that gets fired up once in a while to play '80s vintage games. Sadly, it just doesn't have the memory or storage capacity to run any version of Prime95 (AFAIK, which may not be very far).

Rodrigo

[QUOTE=Rodrigo;223564]How about 190M? [URL]http://www.mersenne.info/?s=100000000&d=1&t=1[/URL][/QUOTE]

That range is about to start being handed out by the server to those who have selected "LMH - TF" on the client.

So, to answer your question in the negative, why don't you do what we've been advising you do for some time, and simply select "LMH - TF" on your client and let the system do what it was designed to do?

[QUOTE=Rodrigo;223564]Which leads me to something I'd been meaning to ask. After all these years and with so many people participating, why is there (apparently) still so much work left undone in the very low ranges? (I.e., [URL]http://www.mersenne.info/?s=0&d=2&t=1[/URL] )

No doubt there is a good reason.[/QUOTE]

Indeed there is...

Remember we told you earlier that the smaller the exponent, the longer TFing takes. Besides, all those low numbers have already been LLed and DCed, so from the project's main objective (finding primes) there is no point in doing additional work on them.

Not to say that you can't, however; and some do and are. But keep in mind it would take your slow machines months to move a single low exponent up by one factor depth, with *very* little chance of finding a factor (because of the PM-1 and possibly EMC factoring attempts done on them).

chalsall,

Thanks for the explanation. I suspected it might have something to do with it taking longer to TF a smaller exponent; just wanted to make sure my suspicion was warranted. :smile:

Yes, LMH-TF does look more and more like the place to go. Things were up in the air for a bit because we were considering OBD, but that's looking less likely now for the P75.

I've signed up the first of my computers to do work, and will be adding each of the others in turn as I get better acquainted with the process. In fact I'll be starting a new thread shortly with a couple of questions about the program settings. But it's already churning away!

Gratefully,

Rodrigo

[quote=Primeinator;223525]I think all of those exponents have been factored past 62 bits (which, if I remember correctly, was the limit of what your oldest machine is capable of).[/quote]
Primeinator,

How can I look up what that limit is?

Thanks!

Rodrigo

There is a little table on here:
[url]http://www.mersenne.org/various/math.php[/url]

I'll reproduce it here slightly edited:[code] Exponents Bit
up to depth
--------- -----
3960000 60
5160000 61
6515000 62
8250000 63
13380000 64
23390000 65
29690000 66
37800000 67
47450000 68
58520000 69
75670000 70
96830000 71[/code]This relationship between log(exponent) and bit depth is roughly linear, with a bit of a kink around 64 and 65 bits.

If you plot the log of the exponent against the bit depth (log to base 2 of the trial factor to depth), you get a nice straight line from a bit depth of 65 onwards.

I added a trend line to that and the equation was this:

ln(E) = 0.2346*D + 1.7171

Where E is the exponent and D is the bit depth.

So using this, it's tentatively possible to extend the table, but whether or not this linear relationship should continue I don't know. Assuming it does, here's what the numbers would be:[code] Exponents Bit
up to depth
--------- -----
23390000 65
29690000 66
37800000 67
47450000 68
58520000 69
75670000 70
96830000 71[color=red]
120640000 72
152530000 73
192860000 74
243860000 75
308340000 76
389860000 77
492940000 78
623280000 79
788070000 80
996440000 81
1259900000 82
1593020000 83
2014220000 84
2546790000 85
3220170000 86
4071590000 87[/color][/code]So it seems as though Operation Billion Digits should be taking numbers to 87 bits, and currently there are 10 (soon to be 13) exponents that are at 81 bits.

Also, the first lot of 100 million digit candidates should be trial factored to 77 bits.

Once again, I do not know if this linear relationship between exponent and trial factor depth should continue. So take it with a pinch of salt until someone who actually knows something about it makes a comment.

[QUOTE=lavalamp;223641]I added a trend line to that and the equation was this:

ln(E) = 0.2346*D + 1.7171

Where E is the exponent and D is the bit depth.

So using this, it's tentatively possible to extend the table, but whether or not this linear relationship should continue I don't know. Assuming it does, here's what the numbers would be:[code] Exponents Bit
up to depth
--------- -----
23390000 65
29690000 66
37800000 67
47450000 68
58520000 69
75670000 70
96830000 71[color=red]
120640000 72
152530000 73
192860000 74
243860000 75
308340000 76
389860000 77
492940000 78
623280000 79
788070000 80
996440000 81
1259900000 82
1593020000 83
2014220000 84
2546790000 85
3220170000 86
4071590000 87[/color][/code]So it seems as though Operation Billion Digits should be taking numbers to 87 bits, and currently there are 10 (soon to be 13) exponents that are at 81 bits.

Also, the first lot of 100 million digit candidates should be trial factored to 77 bits.

Once again, I do not know if this linear relationship between exponent and trial factor depth should continue. So take it with a pinch of salt until someone who actually knows something about it makes a comment.[/QUOTE]
Close enough. The rule of thumb is, doubling of exponent = 3 more bits. If you make it so that the bit depth changes are aligned to FFT size changes, that should be good enough. But since we're looking at only the OBD exponents, the figure 87 bits should be within +/- 1 bit of optimal. Since P-1 is not yet practical on these size exponents, maybe that number can be increased.

[QUOTE=lavalamp;223641]Also, the first lot of 100 million digit candidates should be trial factored to 77 bits.[/QUOTE]That is what George says.:geek:

[QUOTE=axn;223642]If you make it so that the bit depth changes are aligned to FFT size changes, that should be good enough.[/QUOTE]I don't suppose there's a list of these somewhere, or some method to work out what size exponents fit into what size FFT? The benchmark page on mersenne.org only gives exponent ranges for FFT sizes up to 4M, but I know that Prime95 runs to at least 32M.

I wasn't only looking at OBD exponenets by the way, I just gave them a special mention, along with the 100 million digit candidates.

[QUOTE=lavalamp;223654]I don't suppose there's a list of these somewhere, or some method to work out what size exponents fit into what size FFT?[/QUOTE]You will find the FFT size for all exponent ranges up to 596M in the source code. I attach an zipped excel file for those to lazy to look it up.

Jacob

[quote=axn;223642]The rule of thumb is, doubling of exponent = 3 more bits.[/quote]
This can be understood simply:
Doubling the exponent quadruples the time for an LL test,
and halves the number of trial divisors 2kp + 1.

Note also that doubling the exponent multiplies the
computing required to find a Mersenne prime by 8:
the time is 4 times longer and the probability of it
being prime is halved.
If doubling the digits every 3.8 years is to be maintained,
I would guess that GIMPS needs India and/or China to
catch the bug!

David

[QUOTE=S485122;223659]You will find the FFT size for all exponent ranges up to 596M in the source code. I attach an zipped excel file for those to lazy to look it up.[/QUOTE]I downloaded your excel file, so I am clearly lazy.

I also downloaded the source code for Prime95 and looked through the files, none of which are named in such a way as to appear a likely place to contain such information, not that I saw at least. Upon searching the contents of many (MANY) of them, I still have not found it. Since you seem to know where the needle sized array is located in this haystack of 431 files, could you please drop a name?

[QUOTE=lavalamp;223712]I downloaded your excel file, so I am clearly lazy.

I also downloaded the source code for Prime95 and looked through the files, none of which are named in such a way as to appear a likely place to contain such information, not that I saw at least. Upon searching the contents of many (MANY) of them, I still have not found it. Since you seem to know where the needle sized array is located in this haystack of 431 files, could you please drop a name?[/QUOTE]You will find the table in mult.asm
But you could also search the forum : the subject has been treated already.

Jacob

[quote=Primeinator;223525]I think all of those exponents have been factored past 62 bits (which, if I remember correctly, was the limit of what your oldest machine is capable of).[/quote]It's a matter of the default limit in the software, rather than a machine hardware limitation.

If the machine is so old it can't run anything later than Windows for Workgroups 3.11, the default limit in the (ancient) WfW3.11 version of prime95 would apply.

[quote=cheesehead;223738]It's a matter of the default limit in the software, rather than a machine hardware limitation.

If the machine is so old it can't run anything later than Windows for Workgroups 3.11, the default limit in the (ancient) WfW3.11 version of prime95 would apply.[/quote]
cheesehead,

Thanks for this information. It helps me to understand the process much better!

Rodrigo