mersenneforum.org  

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

Reply
 
Thread Tools
Old 2013-06-17, 00:49   #1
Caribou007
 
Jun 2013

23 Posts
Default Overclock Testing Guide?

A quick google search revealed many out of date guides on how to use Prime95 to test CPU stability.

The problem I have found is that the Affinity option is missing from Prime95's Advanced menu.

I have a dual core, socket 939 CPU, the AMD Opteron 180.

Can someone post instructions on how to thoroughly stress test both cores please?
Caribou007 is offline   Reply With Quote
Old 2013-06-17, 02:24   #2
TheMawn
 
TheMawn's Avatar
 
May 2013
East. Always East.

11×157 Posts
Default

As far as I know, you can just run the program and say "Just Stress Testing" or alternatively, pick Torture Test under options.

The distribution of work across the cores should be easy enough. Just choose the same number of threads as you have logical cores on your CPU. Once it's running, check that your CPU is running at 100%. That should suffice.

As for the question of how long is enough, you'll get certain purists who say that 48 hours is minimum, and others who will say that 15 minutes is plenty because you don't want the 48 hours to harm your CPU. What is correct really depends. I run my i5-3570k at 4.6GHz, despite the default being 3.4GHz. When I am watching videos or otherwise internet surfing, Prime95 is running. I only shut if off when I want the absolute snappiest computer (i.e. an intensive 3-D game). During the week, that means almost a solid 120 hours of Prime95.

Is 15 minutes enough? Probably not for any kind of reliability purpose, no. Back when I had 4.8GHz as a target and managed to get it, I would start lowering the voltage a tick at a time just to see how low I could get it. I can't remember exactly, but at something like 1.345V it was running perfectly fine for over an hour, and then at 1.340V about 7 windows services crashed almost immediately, windows Aero kept stopping then starting and stopping, and all my chrome tabs died.

What it comes down to is what you want to do with your processor. If you are doing a two-week (some of the slower processors that still exist today can take 3 months to do a single LL test) job which has ZERO tolerance for hardware errors, you'll want to be certain, and 48 hours is probably good. If you are playing a 3D game, a hardware error is probably going to be completely trivial (one pixel may be rendered in a slightly incorrect shade for one frame).
TheMawn is offline   Reply With Quote
Old 2013-06-17, 18:13   #3
Caribou007
 
Jun 2013

23 Posts
Default

According to this guide,

"If you have a multi-CPU machine then you should run one instance of Prime95 for each CPU. If the cores share one memory controller then you'll probably get maximum stress by running the blend test on one CPU and running a small FFTs test on each of the other cores."

"When you run two or more instances of Prime95 on a multi-CPU machine, each instance doesn't really settle down and run only on one CPU. The instances jump around a little between CPUs. You can see that behaviour if you start up just one instance of Prime95. The CPU Usage History window in the Task Manager will show the program running on both CPUs. You don't end up with one CPU at 100% and the other CPUs at 0%. Prime95 contains a facility to force an instance of the program to run on only one CPU. Go to the main Prime95 window and click the Advanced menu and then select Affinity."

The guide is clearly out of date, and I need to know how to accomplish the same task using the new software.
Caribou007 is offline   Reply With Quote
Old 2013-06-17, 18:40   #4
Batalov
 
Batalov's Avatar
 
"Serge"
Mar 2008
Phi(4,2^7658614+1)/2

24AF16 Posts
Default

Quote:
Originally Posted by Caribou007 View Post
The guide is clearly out of date, and I need to know how to accomplish the same task using the new software.
It is severely out of date. With the new software, you don't have to do anything special (like "-A1" or whatever) to test any and all cores.

Just select Menu -> Options -> "Torture Test", select "Custom" and chose options that you like:
Attached Thumbnails
Click image for larger version

Name:	tort.png
Views:	160
Size:	9.0 KB
ID:	9902  
Batalov is offline   Reply With Quote
Old 2013-06-17, 18:47   #5
Batalov
 
Batalov's Avatar
 
"Serge"
Mar 2008
Phi(4,2^7658614+1)/2

9,391 Posts
Default

(cannot attach two images to one post)

Additionally, you can go to Menu -> Test -> "Worker Windows" and select specific affinity for every worker:
Attached Thumbnails
Click image for larger version

Name:	tort_aff.png
Views:	156
Size:	16.5 KB
ID:	9903  
Batalov is offline   Reply With Quote
Old 2013-06-17, 19:10   #6
Caribou007
 
Jun 2013

108 Posts
Default

If I want to 24-hour test a CPU overclock for stability and do not want the RAM modules, video card or anything else to compromise the test, then should I run the Blend test on one core and the Small FFTs test on the other core, as the guide suggests?* Can each worker be assigned a different test?

*As well as underclocking the RAM.

Last fiddled with by Caribou007 on 2013-06-17 at 19:12
Caribou007 is offline   Reply With Quote
Old 2013-06-17, 19:37   #7
kracker
 
kracker's Avatar
 
"Mr. Meeseeks"
Jan 2012
California, USA

32×241 Posts
Default

Quote:
Originally Posted by Caribou007 View Post
If I want to 24-hour test a CPU overclock for stability and do not want the RAM modules, video card or anything else to compromise the test, then should I run the Blend test on one core and the Small FFTs test on the other core, as the guide suggests?* Can each worker be assigned a different test?

*As well as underclocking the RAM.
I hope you understood what Batalov said...

Anyways for just the CPU the small FFT stress test seems to work well for me in cpu only testing.
kracker is offline   Reply With Quote
Old 2013-06-17, 19:50   #8
Caribou007
 
Jun 2013

810 Posts
Default

Quote:
Originally Posted by kracker View Post
I hope you understood what Batalov said...

Anyways for just the CPU the small FFT stress test seems to work well for me in cpu only testing.
What do you mean you hope I understood it? Of course I did.

However, I still need an answer. Starting a single thread Torture Test after a test is already running does not make Worker #2 start that test. It only makes Worker #1 stop the test it's already running and start the new test instead.

I also set each worker to use a different CPU, however the workload is still split evenly between both cores when running only a single thread. I have to use the Task Manager to set the Affinity to actually get it working on only one core.

Last fiddled with by Caribou007 on 2013-06-17 at 19:59
Caribou007 is offline   Reply With Quote
Old 2013-06-17, 19:56   #9
kracker
 
kracker's Avatar
 
"Mr. Meeseeks"
Jan 2012
California, USA

32×241 Posts
Default

Quote:
Originally Posted by Caribou007 View Post
What do you mean you hope I understood it? Of course I did.

However, I still need an answer.
I thought I answered you, but I guess I didn't, sorry.

Run the Small FFT on both cores for just cpu, both cores on blend if you want RAM testing as well. That guide is quite outdated... You don't have to worry about Affinity or anything like that, it will automatically do everything.
kracker is offline   Reply With Quote
Old 2013-06-17, 20:46   #10
science_man_88
 
science_man_88's Avatar
 
"Forget I exist"
Jul 2009
Dumbassville

203008 Posts
Default

Quote:
Originally Posted by Caribou007 View Post
What do you mean you hope I understood it? Of course I did.
so you understood "severely out of date" and that it may not help you but you seemed to ask if following it made sense to follow it ? that's what I got out of what I saw you post.

Last fiddled with by science_man_88 on 2013-06-17 at 20:46
science_man_88 is offline   Reply With Quote
Old 2013-06-17, 21:05   #11
Caribou007
 
Jun 2013

23 Posts
Default

Um.. maybe you already explained this and I missed it, but..

Starting a single thread Torture Test after a test is already running does not make Worker #2 start that test. It only makes Worker #1 stop the test it's already running and start the new test instead. Can each worker be assigned a different test? Is it better to run one Blend test alongside one Small FFTs test, or two Small FFTs tests?

I also set each worker to use a different CPU, however the workload is still split evenly between both cores when running only a single thread. I have to use the Task Manager to set the Affinity to actually get it working on only one core. So it's not doing everything automatically, and I don't know what to do about that.
Caribou007 is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Can non-K Intel CPUs overclock memory? Prime95 Hardware 5 2015-12-31 22:09
Guide OmbooHankvald Operation Billion Digits 4 2009-08-26 08:18
Guide OmbooHankvald 3*2^n-1 Search 3 2005-07-13 21:03
Why can't I overclock??! db597 Hardware 11 2005-07-05 01:51
Can people overclock caches? clowns789 Hardware 3 2003-07-29 20:27

All times are UTC. The time now is 09:42.

Sun Apr 18 09:42:49 UTC 2021 up 10 days, 4:23, 0 users, load averages: 1.72, 1.53, 1.49

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.