mersenneforum.org  

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

Reply
 
Thread Tools
Old 2012-08-18, 04:33   #1
gamer30
 
Aug 2012

D16 Posts
Talking Stage 2 Memory Settings

Hey guys, I'm just wondering if there's any benefit to setting the memory quite high. I read the readme about it and I used to use:

Day: 512MB
Night: 1024MB

Since I have a lot of memory, I just changed it to:

Day: 2048MB
Night: 4096MB

Is there any benefit to this really? I currently have prime95 running 24 hours/day though I don't always let it run constantly, for instance, if I want to play a game.

Looking forward to your expert responses!
gamer30 is offline   Reply With Quote
Old 2012-08-18, 05:20   #2
Dubslow
Basketry That Evening!
 
Dubslow's Avatar
 
"Bunslow the Bold"
Jun 2011
40<A<43 -89<O<-88

3×2,399 Posts
Default

That's around the point where more memory becomes less and less useful. If you went up to say 4/6 GiB (instead of 2/4) then you might see a few % tops speed improvement, if that. OTOH, .5/1 will give you a 20%+ improvement over .25/.5. Thus 2/4 should be plenty.

Note, of course, that allocated memory only makes a difference for Stage 2 P-1 (not LL tests), but from the thread title, it seems you're aware of that.
Dubslow is offline   Reply With Quote
Old 2012-08-18, 06:07   #3
Xyzzy
 
Xyzzy's Avatar
 
"Mike"
Aug 2002

27·61 Posts
Default

http://www.mersenneforum.org/showthr...16337&p=282335
Xyzzy is offline   Reply With Quote
Old 2012-08-18, 09:17   #4
gamer30
 
Aug 2012

158 Posts
Default

Quote:
Originally Posted by Dubslow View Post
That's around the point where more memory becomes less and less useful. If you went up to say 4/6 GiB (instead of 2/4) then you might see a few % tops speed improvement, if that. OTOH, .5/1 will give you a 20%+ improvement over .25/.5. Thus 2/4 should be plenty.

Note, of course, that allocated memory only makes a difference for Stage 2 P-1 (not LL tests), but from the thread title, it seems you're aware of that.
I actually don't know the difference or even really understand much of what is happening on primes95. I just plug it in, let GIMPS decide what it does (whichever makes more sense setting) and let it go. I just assumed running higher mem for the test was better. THe default is 8MB/8MB!
gamer30 is offline   Reply With Quote
Old 2012-08-18, 10:53   #5
axn
 
axn's Avatar
 
Jun 2003

477710 Posts
Default

Quote:
Originally Posted by gamer30 View Post
Day: 2048MB
Night: 2048MB
It is probably better if you give the same amount to both day & night memory (even if that means lesser amount at night).
axn is offline   Reply With Quote
Old 2012-08-19, 00:08   #6
gamer30
 
Aug 2012

13 Posts
Default

Quote:
Originally Posted by axn View Post
It is probably better if you give the same amount to both day & night memory (even if that means lesser amount at night).
Alright, done. Also, I know this is off topic, but I ran a torture test for 8 hours without error, is that fair that my system is stable? (I know 24 hours is desirable, but I'd rather be running GIMPS.)
gamer30 is offline   Reply With Quote
Old 2012-08-19, 00:10   #7
Dubslow
Basketry That Evening!
 
Dubslow's Avatar
 
"Bunslow the Bold"
Jun 2011
40<A<43 -89<O<-88

3×2,399 Posts
Default

Quote:
Originally Posted by gamer30 View Post
Alright, done. Also, I know this is off topic, but I ran a torture test for 8 hours without error, is that fair that my system is stable? (I know 24 hours is desirable, but I'd rather be running GIMPS.)
If you haven't fiddled with any of the hardware settings, then that's good enough.
Dubslow is offline   Reply With Quote
Old 2012-08-19, 11:13   #8
axn
 
axn's Avatar
 
Jun 2003

17·281 Posts
Default

Quote:
Originally Posted by gamer30 View Post
I ran a torture test for 8 hours without error, is that fair that my system is stable? (I know 24 hours is desirable, but I'd rather be running GIMPS.)
If you want to QA your system while still being productive, run a few doublechecks. That should tell you whether your system is "real world" stable, while still contributing to GIMPS.
axn is offline   Reply With Quote
Old 2012-08-19, 18:44   #9
gamer30
 
Aug 2012

13 Posts
Default

Quote:
Originally Posted by axn View Post
If you want to QA your system while still being productive, run a few doublechecks. That should tell you whether your system is "real world" stable, while still contributing to GIMPS.
I have OC'd it a bit. It seems as long as I don't fiddle too much it's stable. Probably motherboard isn't so great for OC. The doublechecks have me confused, because I disabled error reporting since it was coming up "poor" on the last OC run. When I last enabled round off error checking, I had no issues though I should keep it up longer I guess. I've been running GIMPS for nearly 24 hours without issues, though error checking is off (or disabled.)
gamer30 is offline   Reply With Quote
Old 2012-08-19, 19:18   #10
Dubslow
Basketry That Evening!
 
Dubslow's Avatar
 
"Bunslow the Bold"
Jun 2011
40<A<43 -89<O<-88

3×2,399 Posts
Default

Quote:
Originally Posted by gamer30 View Post
I have OC'd it a bit. It seems as long as I don't fiddle too much it's stable. Probably motherboard isn't so great for OC. The doublechecks have me confused, because I disabled error reporting since it was coming up "poor" on the last OC run. When I last enabled round off error checking, I had no issues though I should keep it up longer I guess. I've been running GIMPS for nearly 24 hours without issues, though error checking is off (or disabled.)
Double checking meaning the GIMPS worktype of LL testing exponents that have already been tested once.

It's good for "real stability testing" because unlike first-time LL, with the DC there's a first residue you can compare yours against. Around 90-95% of the time, your residue should match the first test. (If you can't achieve that success rate, then you've probably pushed the OC too far.)
Dubslow is offline   Reply With Quote
Old 2012-08-19, 19:30   #11
gamer30
 
Aug 2012

11012 Posts
Default

Quote:
Originally Posted by Dubslow View Post
Double checking meaning the GIMPS worktype of LL testing exponents that have already been tested once.

It's good for "real stability testing" because unlike first-time LL, with the DC there's a first residue you can compare yours against. Around 90-95% of the time, your residue should match the first test. (If you can't achieve that success rate, then you've probably pushed the OC too far.)
Pardon me, but I find this confusing. You are talking about enabling "round off checking" and "SUM(INPUTS) error checking" while running GIMPS right?

Currently the "round off" checking is coming up "Round off: 0.0000000000 to 0.xxxxxxxxxx"

Is this normal?
gamer30 is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Memory Settings Fred Software 5 2016-05-03 00:51
Stage 2 Memory Setting - Again Antonio Software 6 2012-09-04 12:48
Memory settings questions Prime95 Hardware 21 2012-02-03 03:34
optimal memory settings for the P-1 stage S485122 Software 16 2007-05-28 12:08
Memory settings and swapping problem MaxP Hardware 3 2003-12-20 05:08

All times are UTC. The time now is 15:27.

Thu Nov 26 15:27:16 UTC 2020 up 77 days, 12:38, 4 users, load averages: 1.69, 1.53, 1.36

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, 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.