mersenneforum.org  

Go Back   mersenneforum.org > New To GIMPS? Start Here! > Information & Answers

Reply
 
Thread Tools
Old 2021-03-12, 22:01   #1
Uncle Lumpy
 
Mar 2017

3×5 Posts
Default Best distributed computing project for CPUs?

Not sure where (which forum) this should go...

I've been running GIMPS since 1996 when a person had to manually contact George to get work assignments. However, I feel as if my CPUs are no longer well suited to searching for Mersenne primes, that GPUs are the place to be these days. I really have no interest in going down the GPU road.

Are there still distributed computing projects for which CPUs have an advantage or, like the dinosaurs, is my CPU-dominated contribution to the world of distributed computing a relic of the past? My Google-foo is not strong on this subject for some reason so I figured the smart folks on these fora would point me in the right direction.

Thanks,
Lumpy
Uncle Lumpy is offline   Reply With Quote
Old 2021-03-12, 22:09   #2
VBCurtis
 
VBCurtis's Avatar
 
"Curtis"
Feb 2005
Riverside, CA

507110 Posts
Default

This reads a bit like an invitation for us to shill our pet projects, so I'll do so:

number-field-sieve (NFS) factoring is an interesting place to be, and most of the tasks involved are CPU-only (or at least commonly done on CPU, in the case of ECM).

Your contribution can be:
1. Fire-and-forget via BOINC client pointed at nfs@home (I suggest the f-small selection in particular if helping we forum denizens makes you happy, or leave default to help all the sub-projects within that factoring effort)

2. Independent work on an aliquot sequence, where each term is the sum of the previous term's factors. This is best if you have interest in learning how the individual factoring tools work. YAFU is the software wrapper that handles most of the work.

3. You can run ECM work, either on a number of the size that might get factored by NFS in the future, or on mersenne exponents. See the "getting under 20M unfactored" thread for details on that Mersenne sub-project's work.

If factoring isn't your thing, running LLR or PFGW on smaller exponents to search for primes is still done on CPU alone (LLR has a GPU version, but it's not a great use of GPU resources and CPU version is more efficient). Conjectures R Us and the riesel prime search subforums are popular examples of teams organizing work, but the prime search is mostly independent and at your own pace.
VBCurtis is offline   Reply With Quote
Old 2021-03-12, 22:45   #3
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101Γ—103 Posts

2·5,059 Posts
Default

There is still a huge back log of LL-DC's that need to get done. (These can also be done as a PRP with a cert.) CPU's are still worthwhile in this range. And even though GPU's may be vastly more powerful, all cycles contribute.

Also, there is the newish task of doing a PRP test on the cofactors of Mersenne numbers. This allows us to know if the cofactor is prime (and thus the Mersenne number is fully factored) or not. Right now those are being done in the 10.8M range and up. So, an older CPU can turn these in frequently. It is not as glamorous as finding a new prime, but an old CPU can continue to contribute.
Uncwilly is offline   Reply With Quote
Old 2021-03-13, 07:44   #4
pinhodecarlos
 
pinhodecarlos's Avatar
 
"Carlos Pinho"
Oct 2011
Milton Keynes, UK

3×1,663 Posts
Default

I’m on 1) and sometimes I help GIMPS on P-1 but I’m on the same boat as you are. My laptop is too slow for first wave. Happy to support you on 1) at NFS@Home.
pinhodecarlos is online now   Reply With Quote
Old 2021-03-13, 14:42   #5
Uncle Lumpy
 
Mar 2017

3×5 Posts
Default

Thanks for the responses, folks. For now, I've switched my CPUs to working on the DC backlog while I investigate nfs@home. I did fiddle with the BOINC client 10 years ago or so during a brief SETI@home phase, but always kept my toes in the GIMPS waters.

Thanks!
Lumpy
Uncle Lumpy is offline   Reply With Quote
Old 2021-03-29, 08:20   #6
M0CZY
 
M0CZY's Avatar
 
May 2005
England, UK

22×31 Posts
Default

I've got an old laptop that is very slow by modern standards (Intel Pentium M Mobile @ 2.26 GHz), and it has 2 GB of RAM.

Which of these ECM subprojects have the shortest assignments?

5 - ECM for first factors of Mersenne numbers
8 - ECM on Mersenne cofactors
M0CZY is offline   Reply With Quote
Old 2021-03-29, 14:05   #7
xilman
Bamboozled!
 
xilman's Avatar
 
"π’‰Ίπ’ŒŒπ’‡·π’†·π’€­"
May 2003
Down not across

1102610 Posts
Default

Quote:
Originally Posted by Uncle Lumpy View Post
Not sure where (which forum) this should go...

Are there still distributed computing projects for which CPUs have an advantage or, like the dinosaurs, is my CPU-dominated contribution to the world of distributed computing a relic of the past?
Undoubtedly so, as long as you don't mind working on non-Mersenne numbers.

Integer factorization, for instance. My ECMNET server is plugging away after many years. There are some Number Field Sieve collaborations too.

If you would like to help me out, the minimum you need to know is one line in a configuration file.

server=100:a:83.217.167.177:8194

I can provide vastly more information to get you up and running relatively easily.

Paul
xilman is offline   Reply With Quote
Old 2021-03-29, 14:59   #8
Uncle Lumpy
 
Mar 2017

3·5 Posts
Default

Quote:
Originally Posted by xilman View Post
Undoubtedly so, as long as you don't mind working on non-Mersenne numbers.

Integer factorization, for instance. My ECMNET server is plugging away after many years. There are some Number Field Sieve collaborations too.

If you would like to help me out, the minimum you need to know is one line in a configuration file.

server=100:a:83.217.167.177:8194

I can provide vastly more information to get you up and running relatively easily.

Paul
Paul,

Thanks for the response. I'm presently pumping out 3-4 LL double checks per day (a drop in the bucket as to what is needed?) but would be interested in hearing more about ECMNET. Not sure I'll switch over (one big thing that GIMPS provides is a "set it and forget it" client), but I would like to hear more.

Best,
Lumpy
Uncle Lumpy is offline   Reply With Quote
Old 2021-03-29, 16:23   #9
Prime95
P90 years forever!
 
Prime95's Avatar
 
Aug 2002
Yeehaw, FL

1E0316 Posts
Default

Quote:
Originally Posted by M0CZY View Post
Which of these ECM subprojects have the shortest assignments?

5 - ECM for first factors of Mersenne numbers
8 - ECM on Mersenne cofactors
I suspect 8 would be faster and more likely to have occasional successes. 5 is a little more noteworthy when you have a success. Your machine is well suited to either work preference.
Prime95 is offline   Reply With Quote
Old 2021-03-29, 17:41   #10
M0CZY
 
M0CZY's Avatar
 
May 2005
England, UK

12410 Posts
Default

Quote:
I suspect 8 would be faster and more likely to have occasional successes.
Thank you for that information. I will set that work preference in my laptop's client.

It's just a shame that you can't specify 'certification' as a work type, although I suspect that there might not be enough certification work for everyone who would like to do it.

It's a pity, because it would give slow computers a real chance to take part in this project in a meaningful way.
M0CZY is offline   Reply With Quote
Old 2021-04-26, 06:35   #11
bur
 
bur's Avatar
 
Aug 2020
79*6581e-4;3*2539e-3

2×211 Posts
Default

Interesting that you perceive it that way, I always thought CPUs are suprior to GPU when it comes to prime hunting and only specialized applications such as TF or primality of generalized Fermat numbers is for GPU. Deterministic prime tests for most types of primes such as with LLR and pfgw are most efficient on CPU to my knowledge.

If you want a fire-and-forget for prime hunting outside GIMPS I think your best bet is Primegrid. It's a BOINC project that does primality tests on a large variety of prime numbers. You can choose what you want to test and then just leave it running. It's very userfriendly due to it using the BOINC framework. And with the many subprojects, if you get bored by one search you can switch to another.

So LL tests on Mersenne numbers are more efficient on GPU than on CPU?
bur is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
AlphaZero style distributed computing project for Go and Chess MooMoo2 Lounge 5 2019-05-28 13:22
distributed computing project to prove/disprove Riemann Hypothesis ssybesma Other Mathematical Topics 10 2018-12-18 13:26
GNFS distributed computing andi314 Factoring 50 2005-05-11 00:56
A distributed-computing project to optimize GIMPS FFT? Genetic algorithms GP2 Software 10 2003-12-09 20:41
The difference between P2P and distributed computing and grid computing GP2 Lounge 2 2003-12-03 14:13

All times are UTC. The time now is 08:32.


Fri Dec 3 08:32:32 UTC 2021 up 133 days, 3:01, 0 users, load averages: 1.13, 1.50, 1.81

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.