mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > Software

Reply
 
Thread Tools
Old 2006-12-23, 23:08   #1
tha
 
tha's Avatar
 
Dec 2002

829 Posts
Default Renaming Prime95

I just finished assembling a new pc for under the X-mas tree. My 866 MHz Pentium III machine will be replaced by a Core 2 Duo E6600. I nicely managed to skip the flawed Pentium IV design era.

So for the first time I installed two copies of Prime95, each in it's own directory, and set the affinity to respectively core 0 and core 1. As most people do, I ran a torture test on both instances and watched the CPU activity reaching an unexpected value less than 100%. After browsing the forum a bit further I found out torture test for multi core processors are not supported. Running two real exponents on both instances worked perfectly well and gave the desired 100% CPU usage value. (And a rise in temperature from 60 to 70 degrees Celsius.)

During the search for extra information on the internet it struck me how many pages are available advising overclockers to download a copy of prime95 and run it for testing purposes. No wonder so many assignment are abandoned right away. The 'just stress testing' option at the beginning helps, but only for single core pc's. Specific programs designed for stress testing using the Prime95 source code available here and there are even better.

But I think we will never get rid of all the people out there who download Prime95 as a matter of habit or after having read the advise to do so on the many available pages. If and when the v5 server becomes available it might be a good idea to rename the client to something including the name 'Mersenne' and without the 95 numeral. And then create a new page on the Mersenne website where it will be available for download. Right now it is in the directory 'gimps', where we should leave the v4 Prime95 client but not link to it from the rest of the site. This looks to me the best option to spin off the overclockers further away from the GIMPS server.
tha is offline   Reply With Quote
Old 2006-12-24, 01:24   #2
retina
Undefined
 
retina's Avatar
 
"The unspeakable one"
Jun 2006
My evil lair

2×23×137 Posts
Default

I don't see any problem with "overclockers" taking an assignment or two. Even if they abandon them shortly afterwards it doesn't matter, the exponents get reassigned pretty quickly (~60 days). It's not like there is a dearth of available exponents, there are plenty to go around.

Who knows, maybe a small percentage will keep it running and complete a few tests, it helps towards the overall goal.
retina is offline   Reply With Quote
Old 2006-12-24, 12:27   #3
Mini-Geek
Account Deleted
 
Mini-Geek's Avatar
 
"Tim Sorbera"
Aug 2006
San Antonio, TX USA

2·5·7·61 Posts
Default

Doesn't Prime95 ask if you want to join GIMPS or just torture test? Even if someone accidentally joins then stops later on, as retina pointed out, it gets reassigned rather quickly.
Mini-Geek is offline   Reply With Quote
Old 2006-12-25, 22:19   #4
tha
 
tha's Avatar
 
Dec 2002

829 Posts
Default

Quote:
Originally Posted by Mini-Geek View Post
Doesn't Prime95 ask if you want to join GIMPS or just torture test? Even if someone accidentally joins then stops later on, as retina pointed out, it gets reassigned rather quickly.
Over 50% of all server assignments are done for stress testing and abandoned right away. These assignments last about 80 days on average. About a quarter of all exponents go through this stage twice and an eighth three times and so on. Of course there are enough exponents available, but this way our not so fast processors have to churn on rather large exponents which could better be done when they are replaced by faster ones. And I like a clean database which has ranges completed rather then many individual exponents.
tha is offline   Reply With Quote
Old 2006-12-25, 22:57   #5
retina
Undefined
 
retina's Avatar
 
"The unspeakable one"
Jun 2006
My evil lair

2·23·137 Posts
Default

Quote:
Originally Posted by tha
Over 50% of all server assignments are done for stress testing and abandoned right away.
How do you know that? Did you survey all those people to ask what they were doing? It could be good intentioned people that forget about prime95 and then reformat their machines. Also, how would testing an exponent help to test their system? They would not know the correct result to compare. And if they don't wait for the correct result then what was the point of running it anyway? Surely the stress test gives a more reliable result and gives it faster.
Quote:
Originally Posted by tha
These assignments last about 80 days on average.
So? The project has been running for >10 years.
Quote:
Originally Posted by tha
About a quarter of all exponents go through this stage twice and an eighth three times and so on.
Okay, I'll take your word for it.
Quote:
Originally Posted by tha
Of course there are enough exponents available, but this way our not so fast processors have to churn on rather large exponents which could better be done when they are replaced by faster ones.
No matter how one distributes the exponents the work still has to be done. It makes no difference, the total number of FLOPS still needs to be done whether one does the "faster ones" first or not.
Quote:
Originally Posted by tha
And I like a clean database which has ranges completed rather then many individual exponents.
Now I think I see the main reason for your post. Yeah the distribution looks a bit messy, but that is only on the surface, underneath there is still a lot of good work being done.

Last fiddled with by retina on 2006-12-25 at 22:59
retina is offline   Reply With Quote
Old 2006-12-30, 05:38   #6
drew
 
drew's Avatar
 
Jun 2005

2·191 Posts
Default

Quote:
Originally Posted by Mini-Geek View Post
Doesn't Prime95 ask if you want to join GIMPS or just torture test? Even if someone accidentally joins then stops later on, as retina pointed out, it gets reassigned rather quickly.
...if I may go off-topic for a sec.

Mini-geek, I really like the avatar, but have you noticed that the light source (as perceived by the specular reflection on the steel marbles) appears to shift sides as the animation shifts from left to right? It's apparent that the animation was constructed by flipping the first half of the animation to act as the second half.
drew is offline   Reply With Quote
Old 2006-12-30, 12:45   #7
Mini-Geek
Account Deleted
 
Mini-Geek's Avatar
 
"Tim Sorbera"
Aug 2006
San Antonio, TX USA

10000101011102 Posts
Default

Quote:
Originally Posted by drew View Post
...if I may go off-topic for a sec.

Mini-geek, I really like the avatar, but have you noticed that the light source (as perceived by the specular reflection on the steel marbles) appears to shift sides as the animation shifts from left to right? It's apparent that the animation was constructed by flipping the first half of the animation to act as the second half.
I got it off the claim your avatar thread. I guess I've noticed it but now that you've pointed it out, I'll have to fix it or it will drive me nuts. Well, I'm off to fix it...
EDIT: Ok, it's fixed.

Last fiddled with by Mini-Geek on 2006-12-30 at 13:04
Mini-Geek is offline   Reply With Quote
Old 2006-12-30, 14:28   #8
drew
 
drew's Avatar
 
Jun 2005

2·191 Posts
Default

Quote:
Originally Posted by Mini-Geek View Post
I got it off the claim your avatar thread. I guess I've noticed it but now that you've pointed it out, I'll have to fix it or it will drive me nuts. Well, I'm off to fix it...
EDIT: Ok, it's fixed.
Looks good!
drew is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Renaming GIMPS? Madpoo Lounge 63 2019-05-23 22:07

All times are UTC. The time now is 01:28.


Wed Dec 8 01:28:38 UTC 2021 up 137 days, 19:57, 0 users, load averages: 1.63, 1.40, 1.27

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.