Retrieve work that makes most sense
 

I have an old P2 450MHz lying around, so I decided to run GIMPS on it. It is set to use Primenet, and to retrieve work that makes most sense. But I just noticed that it retrieved LL work, and the estimated time to completion is 337 days! I'm not sure if this old machine is even going to last that long before it breaks down.

My questions:
1. why has it been assigned work that's unreasonable for it to solve
2. what's the best work to do on an old P2 (trial factoring?)

I guess George was going to wait until v5 before they changed the "do whatever makes most sense". I am assuming that it was given a double check to do. Doublechecking has come a long way recently, so yes, it will take a long time to do.

If you don't want to wait that long, just unreserve the exponent (rather than waiting for it to expire) and tell it to get factoring assignments.

I have a P3 550 and I run factoring assignments on it. It takes about a week per exponent or so.

Just be warned that you will never find a mersenne prime by factoring. You probably already knew this.... but just incase you didn't.....

[QUOTE=db597]I have an old P2 450MHz lying around, so I decided to run GIMPS on it. [ ... ] what's the best work to do on an old P2 (trial factoring?)[/QUOTE]

If you don't care about the credit, but still want to help the GIMPS project in an interesting way, you might consider the Lone Mersenne Hunters (LMH) project. I'm doing prefactoring up to 2^62 on an old 700 MHz Celeron with 64MB RAM running Linux. Occasionally, I find several factors in a day :cat:

regards, Leif.

Thanks for the very quick replies Dave and Leif.

I've unreserved the exponent and set the old P2 on factoring assignments. The time to do the factoring assignment it is now given by PrimeNet is 2 weeks. Much better! :)

The idea of LMH is an interesting one, and I did consider it for a while. But it doesn't use PrimeNet and I'm not so keen on manually assigning work to the machine. Thanks for the suggestion nevertheless.

Last question while I've got your attention - what's the best version of Prime95 to run on a P2? I've got it running the latest client (24.11 I believe), but that's probably not the most ideal.

[QUOTE=db597]Last question while I've got your attention - what's the best version of Prime95 to run on a P2? I've got it running the latest client (24.11 I believe), but that's probably not the most ideal.[/QUOTE]

Generally, the newer version you've got, the better. For each version, the code will be both more stable and more optimized. Besides, Prime95 auto-senses what kind of CPU it's running on and uses highly optimized code for each major CPU type. Code for older processors doesn't get thrown out.

regards, Leif.

[QUOTE=leifbk]Generally, the newer version you've got, the better. For each version, the code will be both more stable and more optimized. Besides, Prime95 auto-senses what kind of CPU it's running on and uses highly optimized code for each major CPU type. Code for older processors doesn't get thrown out.

regards, Leif.[/QUOTE]

I think version 23.2 best fits his needs.

E.T.

Thanks once again for the replies. I checked in mersenne.org/gimps. There's a 23.5, but no 23.2.

23.2, 23.5, 24.11 all have the same x87 trial factoring code.

24.11 shouldnot have given you a double-check assignment. Does Options/CPU properly identify your machine? Does prime.log have any error codes stating that the server ran out of factoring assignments?

[QUOTE=Prime95]24.11 should not have given you a first-time assignment. [/QUOTE]

Bug found and fixed. Thanks for finding it.

Yep, in Prime95 the machine is correct identified as a Pentium II or Pentium II Xeon. Cache sizes and clock speed are also identified correctly. In the local.ini file it says that the oldcputype is 9.

The log files do not indicate that the server has run out of factoring assignments, only says it is retrieving an exponent (like usual on my P4s for LL).

Just read that you've found the bug. So I guess no worries about it now. Cheers for fixing. Very fast response time (even before I could reply to your earlier post). Keep up the good work!

Just out of curiousity, what are the current machine boundaries for TF, DC and LL tests? Had a short look at whatsnew.txt, but don't see changes there.