mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > Hardware > GPU Computing

Reply
 
Thread Tools
Old 2013-09-26, 06:38   #1
Mark Rose
 
Mark Rose's Avatar
 
"/X\(โ€˜-โ€˜)/X\"
Jan 2013

3·11·89 Posts
Default Best use for my cards?

I recently purchased a GeForce GTX 760. It gets about 232 GHz/days/day running mfaktc-0.20 (yes, I've submitted benchmarks). This is in my desktop, but I don't do much 3d stuff day to day.

I also have a GeForce GT 430 (which gets about 52 GHz/days/day).

What would be the best use for these cards? The software must run on 64-bit Linux, and should be automatic as much as possible (especially for the GT430 card, which I will put in a box that current runs only mprime).
Mark Rose is offline   Reply With Quote
Old 2013-09-26, 07:25   #2
xilman
Bamboozled!
 
xilman's Avatar
 
"๐’‰บ๐’ŒŒ๐’‡ท๐’†ท๐’€ญ"
May 2003
Down not across

3·5·743 Posts
Default

Quote:
Originally Posted by Mark Rose View Post
I recently purchased a GeForce GTX 760. It gets about 232 GHz/days/day running mfaktc-0.20 (yes, I've submitted benchmarks). This is in my desktop, but I don't do much 3d stuff day to day.

I also have a GeForce GT 430 (which gets about 52 GHz/days/day).

What would be the best use for these cards? The software must run on 64-bit Linux, and should be automatic as much as possible (especially for the GT430 card, which I will put in a box that current runs only mprime).
How much effort do you want to put into it?

For instance, GMP-ECM runs stage 1 very effectively on a GPU but stage 2 has to be completed on a cpu. I could very easily provide you with a large number of integers for which you could compute stage 1 and then either complete yourself or return the save files to me for completion. A little bit of scripting and the occasional (weekly or monthly) email and it should run unattended.

If you want to get your hands dirty in the short term in return for completely automatic processing afterwards, perhaps we could work together on an ECMNET client which runs stage 1 on the GPU. It's something I've wanted for quite a while and would be widely useful.

Paul
xilman is offline   Reply With Quote
Old 2013-09-26, 16:13   #3
TheMawn
 
TheMawn's Avatar
 
May 2013
East. Always East.

11·157 Posts
Default

Universally, I think Trial Factoring will get you the most GHz-Days and we aren't by any stretch too far ahead of the LL-front to be adding more firepower to TF. You CAN do P-1, ECM and LL on your GPU but the question is how much productivity are you willing to sacrifice to work on a slightly unconventional task.

You should probably investigate the ECM project and see if that suits you. My personal preference is GHz-Days maximization and I'd rather get trial factoring away from CPUs because GPUs do it so much better.
TheMawn is offline   Reply With Quote
Old 2013-09-27, 03:18   #4
Mark Rose
 
Mark Rose's Avatar
 
"/X\(โ€˜-โ€˜)/X\"
Jan 2013

293710 Posts
Default

Quote:
Originally Posted by xilman View Post
How much effort do you want to put into it?

For instance, GMP-ECM runs stage 1 very effectively on a GPU but stage 2 has to be completed on a cpu. I could very easily provide you with a large number of integers for which you could compute stage 1 and then either complete yourself or return the save files to me for completion. A little bit of scripting and the occasional (weekly or monthly) email and it should run unattended.

If you want to get your hands dirty in the short term in return for completely automatic processing afterwards, perhaps we could work together on an ECMNET client which runs stage 1 on the GPU. It's something I've wanted for quite a while and would be widely useful.

Paul
Hi Paul,

I've never looked into ECM until now. I don't have much of a math background. I don't know if I would be the best person to assist with that project :/
Mark Rose is offline   Reply With Quote
Old 2013-09-27, 03:33   #5
Mark Rose
 
Mark Rose's Avatar
 
"/X\(โ€˜-โ€˜)/X\"
Jan 2013

3·11·89 Posts
Default

Quote:
Originally Posted by TheMawn View Post
Universally, I think Trial Factoring will get you the most GHz-Days and we aren't by any stretch too far ahead of the LL-front to be adding more firepower to TF. You CAN do P-1, ECM and LL on your GPU but the question is how much productivity are you willing to sacrifice to work on a slightly unconventional task.

You should probably investigate the ECM project and see if that suits you. My personal preference is GHz-Days maximization and I'd rather get trial factoring away from CPUs because GPUs do it so much better.
Is it best to get manual assignments from mersenne.org or to figure out GPU72?
Mark Rose is offline   Reply With Quote
Old 2013-09-27, 03:37   #6
LaurV
Romulan Interpreter
 
LaurV's Avatar
 
"name field"
Jun 2011
Thailand

987010 Posts
Default

@op: Download MISFIT. Read a bit about, possible pick gpu72 as a provider for your exponents, set it into MISFIT and forget it. That is the easiest and makes the most use of your card. You can not find primes by doing that, however (i.e. no money rewards for you ), but you will help the project.

Quote:
Originally Posted by xilman View Post
If you want to get your hands dirty in the short term in return for completely automatic processing afterwards, perhaps we could work together on an ECMNET client which runs stage 1 on the GPU. It's something I've wanted for quite a while and would be widely useful.

Paul
I may be taking up on this offer... however my time is very limited...
LaurV is offline   Reply With Quote
Old 2013-09-27, 03:47   #7
Batalov
 
Batalov's Avatar
 
"Serge"
Mar 2008
Phi(4,2^7658614+1)/2

2×37×131 Posts
Default

Just for the sake of variety, you could try the world record GFN project: http://primegrid.com/stats_genefer.php (Select N=22 ... or N=20 is you don't want the world record). Setup is easy: BOINC, attach, select subproject, and you are done. You will also get the validation of your cards' stability. Or instability, as they case might be.
Batalov is offline   Reply With Quote
Old 2013-09-27, 04:26   #8
Mark Rose
 
Mark Rose's Avatar
 
"/X\(โ€˜-โ€˜)/X\"
Jan 2013

1011011110012 Posts
Default

Quote:
Originally Posted by LaurV View Post
@op: Download MISFIT. Read a bit about, possible pick gpu72 as a provider for your exponents, set it into MISFIT and forget it. That is the easiest and makes the most use of your card. You can not find primes by doing that, however (i.e. no money rewards for you ), but you will help the project.
It seems MISFIT is Windows only?

I don't care much about the monetary rewards.
Mark Rose is offline   Reply With Quote
Old 2013-09-27, 04:38   #9
Mark Rose
 
Mark Rose's Avatar
 
"/X\(โ€˜-โ€˜)/X\"
Jan 2013

B7916 Posts
Default

Quote:
Originally Posted by Batalov View Post
Just for the sake of variety, you could try the world record GFN project: http://primegrid.com/stats_genefer.php (Select N=22 ... or N=20 is you don't want the world record). Setup is easy: BOINC, attach, select subproject, and you are done. You will also get the validation of your cards' stability. Or instability, as they case might be.
I could do that. Perhaps I'll try this weekend.

Right now I'm running about 1000 total TFs from 66-70 to 72 manually checked out from mersenne.org.

Last fiddled with by Mark Rose on 2013-09-27 at 04:41
Mark Rose is offline   Reply With Quote
Old 2013-09-27, 18:59   #10
garo
 
garo's Avatar
 
Aug 2002
Termonfeckin, IE

22·691 Posts
Default

Mark I would recommend using GPU72 to get exponents. We are currently doing TF to 74 bits as that helps the project more than just 72 bits. We are a bit resource constrained so another GPU doing TF would really help. Also factoring to 74 bits means you will spend 4 times as long on each exponent and thus will need to get assignments less often. GPU72 figures out which exponents need doing most and helps optimise your resources for GIMPS. If MISFIT doesn't work for you, you should look at https://github.com/teknohog/primetools (written by forum contributor teknohog)
which uses python and does automatic work fetching and result submission. You will need to install python on your machine but if you are running Linux I assume that would be easy enough for you. Welcome and best of luck.

Last fiddled with by garo on 2013-09-27 at 19:05
garo is offline   Reply With Quote
Old 2013-10-01, 07:27   #11
Mark Rose
 
Mark Rose's Avatar
 
"/X\(โ€˜-โ€˜)/X\"
Jan 2013

3×11×89 Posts
Default

Quote:
Originally Posted by garo View Post
Mark I would recommend using GPU72 to get exponents. We are currently doing TF to 74 bits as that helps the project more than just 72 bits. We are a bit resource constrained so another GPU doing TF would really help. Also factoring to 74 bits means you will spend 4 times as long on each exponent and thus will need to get assignments less often. GPU72 figures out which exponents need doing most and helps optimise your resources for GIMPS. If MISFIT doesn't work for you, you should look at https://github.com/teknohog/primetools (written by forum contributor teknohog)
which uses python and does automatic work fetching and result submission. You will need to install python on your machine but if you are running Linux I assume that would be easy enough for you. Welcome and best of luck.
Setup primetools tonight. Seems to be working. Should take me about 2h45m to TF a 74 bit on my main box, so I'll see how it goes overnight. I also discovered I had another GT 430 card sitting around. Now to find a machine to plug it into...
Mark Rose is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Prime95 and graphics cards keithschmidt Information & Answers 45 2016-09-10 10:08
Too many video cards cardogab7341 GPU Computing 8 2014-12-26 03:32
Cudalucas on cheap cards PageFault GPU Computing 13 2012-05-04 13:44
credit cards- should they be illegal ? science_man_88 Science & Technology 36 2010-11-26 11:27
Using the GPU on video cards for GIMPS? dans Hardware 4 2004-05-09 23:58

All times are UTC. The time now is 04:05.


Tue Jan 25 04:05:49 UTC 2022 up 185 days, 22:34, 0 users, load averages: 0.94, 1.29, 1.37

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2022, 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.

โ‰  ยฑ โˆ“ รท ร— ยท โˆ’ โˆš โ€ฐ โŠ— โŠ• โŠ– โŠ˜ โŠ™ โ‰ค โ‰ฅ โ‰ฆ โ‰ง โ‰จ โ‰ฉ โ‰บ โ‰ป โ‰ผ โ‰ฝ โŠ โŠ โŠ‘ โŠ’ ยฒ ยณ ยฐ
โˆ  โˆŸ ยฐ โ‰… ~ โ€– โŸ‚ โซ›
โ‰ก โ‰œ โ‰ˆ โˆ โˆž โ‰ช โ‰ซ โŒŠโŒ‹ โŒˆโŒ‰ โˆ˜ โˆ โˆ โˆ‘ โˆง โˆจ โˆฉ โˆช โจ€ โŠ• โŠ— ๐–• ๐–– ๐–— โŠฒ โŠณ
โˆ… โˆ– โˆ โ†ฆ โ†ฃ โˆฉ โˆช โŠ† โŠ‚ โŠ„ โŠŠ โŠ‡ โŠƒ โŠ… โŠ‹ โŠ– โˆˆ โˆ‰ โˆ‹ โˆŒ โ„• โ„ค โ„š โ„ โ„‚ โ„ต โ„ถ โ„ท โ„ธ ๐“Ÿ
ยฌ โˆจ โˆง โŠ• โ†’ โ† โ‡’ โ‡ โ‡” โˆ€ โˆƒ โˆ„ โˆด โˆต โŠค โŠฅ โŠข โŠจ โซค โŠฃ โ€ฆ โ‹ฏ โ‹ฎ โ‹ฐ โ‹ฑ
โˆซ โˆฌ โˆญ โˆฎ โˆฏ โˆฐ โˆ‡ โˆ† ฮด โˆ‚ โ„ฑ โ„’ โ„“
๐›ข๐›ผ ๐›ฃ๐›ฝ ๐›ค๐›พ ๐›ฅ๐›ฟ ๐›ฆ๐œ€๐œ– ๐›ง๐œ ๐›จ๐œ‚ ๐›ฉ๐œƒ๐œ— ๐›ช๐œ„ ๐›ซ๐œ… ๐›ฌ๐œ† ๐›ญ๐œ‡ ๐›ฎ๐œˆ ๐›ฏ๐œ‰ ๐›ฐ๐œŠ ๐›ฑ๐œ‹ ๐›ฒ๐œŒ ๐›ด๐œŽ ๐›ต๐œ ๐›ถ๐œ ๐›ท๐œ™๐œ‘ ๐›ธ๐œ’ ๐›น๐œ“ ๐›บ๐œ”