mersenneforum.org  

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

Reply
 
Thread Tools
Old 2009-06-29, 22:46   #1
GARYP166
 
GARYP166's Avatar
 
Feb 2009

438 Posts
Default LL tests running at different speeds

I upgraded my PC last week. Prime 95 seems to have detected this, because I am now testing two numbers. But they never run at the same speed.

I then tried to copy and paste from the screen displays, and found that I have to copy by hand.

[Jun 29 18:17] Iteration 30130000 / 40962983 [73.55%] Per iteration time 0.158 sec
[Jun 29 18:36] Iteration 30140000 / 40962983 [73.57%] Per iteration time 0.113 sec
[Jun 29 19:00] Iteration 30150000 / 40962983 [73.60%] Per iteration time 0.140 sec
[Jun 29 20:46] Iteration 30160000 / 40962983 [73.62%] Per iteration time 0.491 sec
[Jun 29 21:08] Iteration 30170000 / 40962983 [73.65%] Per iteration time 0.129 sec

[Jun 29 18:03] Iteration 2730000 / 39815987 [6.85%] Per iteration time 0.110 sec
[Jun 29 18:24] Iteration 2740000 / 39815987 [6.88%] Per iteration time 0.122 sec
[Jun 29 18:50] Iteration 2750000 / 39815987 [6.90%] Per iteration time 0.159 sec
[Jun 29 19:13] Iteration 2760000 / 39815987 [6.93%] Per iteration time 0.137 sec
[Jun 29 19:28] Iteration 2770000 / 39815987 [6.95%] Per iteration time 0.088 sec

I am happy with the program speeding up or slowing down depending on what else I am doing, but why are the two primes not keeping in step with each other?

Gary
GARYP166 is offline   Reply With Quote
Old 2009-06-29, 23:19   #2
cheesehead
 
cheesehead's Avatar
 
"Richard B. Woods"
Aug 2002
Wisconsin USA

11110000011002 Posts
Default

Quote:
Originally Posted by GARYP166 View Post
I am happy with the program speeding up or slowing down depending on what else I am doing, but why are the two primes not keeping in step with each other?
The two exponents' tests are running in different cores. The "what else [you are] doing" runs on various cores at various times, so there's no lock-step.

Suppose 40962983 is being tested on core 0 and 39815987 is being tested on core 1.

Then when you start up Windows Explorer, for example, the OS may schedule it to run on core 0. Core 1 would be unaffected and the 39815987 test will proceed at the same speed as before, but the prime95 test of 40962983 on core 0 will be slowed down by interruptions for the higher-priority time slices given to Windows Explorer on that core.

Last fiddled with by cheesehead on 2009-06-29 at 23:21
cheesehead is offline   Reply With Quote
Old 2009-06-30, 10:09   #3
GARYP166
 
GARYP166's Avatar
 
Feb 2009

5×7 Posts
Default

Thanks
GARYP166 is offline   Reply With Quote
Old 2009-06-30, 12:36   #4
Mini-Geek
Account Deleted
 
Mini-Geek's Avatar
 
"Tim Sorbera"
Aug 2006
San Antonio, TX USA

426910 Posts
Default

Try finding the times they get when nothing else is using the CPU. Set the "Iterations between screen outputs" (in Options > Preferences) to something low like 10. Bring up Task Manager to check that Prime95 is getting all of your CPU. If they're about the same when it's like that, then it is just that other tasks get sent to one core more than the other. If not, then something else is going on. Either way, you can set that number back to 10000 so it doesn't flood your screen all the time.
Are the two numbers using the same FFT length? It says the FFT length used when you start/resume the number, e.g. "Resuming primality test of M40962983 using FFT length 2560K". Your two candidates are near an FFT cutoff. I don't think they should be using different FFT sizes unless your CPU doesn't support SSE2, in which case it wouldn't have two cores in the first place. (if this doesn't make sense, don't worry about it, just post those FFT size numbers from Prime95 )
Do you know if your CPU has two separate cores or if it is hyperthreaded? It might be related to that.
Mini-Geek is offline   Reply With Quote
Old 2009-07-01, 16:38   #5
starrynte
 
starrynte's Avatar
 
Oct 2008
California

22×59 Posts
Default

Also, go to Edit -> Copy Window to copy the text from whichever window is selected at the time.
starrynte is offline   Reply With Quote
Old 2009-07-04, 10:53   #6
henryzz
Just call me Henry
 
henryzz's Avatar
 
"David"
Sep 2007
Cambridge (GMT/BST)

22×1,481 Posts
Default

in general i find on windows that a low priority job running on core 0 will lose more time than any other core
henryzz is offline   Reply With Quote
Old 2009-07-04, 19:45   #7
cheesehead
 
cheesehead's Avatar
 
"Richard B. Woods"
Aug 2002
Wisconsin USA

1E0C16 Posts
Default

Quote:
Originally Posted by henryzz View Post
in general i find on windows that a low priority job running on core 0 will lose more time than any other core
I agree.
cheesehead is offline   Reply With Quote
Old 2009-07-04, 20:15   #8
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

52×401 Posts
Default

Quote:
Originally Posted by henryzz View Post
in general i find on windows that a low priority job running on core 0 will lose more time than any other core
There are times that I do things that tie up a core with a browser. Core 1 is the one that suffers, it can virtually stop dead, while core 0 pumps right along.
Uncwilly is online now   Reply With Quote
Old 2009-07-04, 23:35   #9
henryzz
Just call me Henry
 
henryzz's Avatar
 
"David"
Sep 2007
Cambridge (GMT/BST)

592410 Posts
Default

Quote:
Originally Posted by Uncwilly View Post
There are times that I do things that tie up a core with a browser. Core 1 is the one that suffers, it can virtually stop dead, while core 0 pumps right along.
weird
i must point out that i have found this with a quad not a duo
henryzz is offline   Reply With Quote
Old 2009-07-12, 21:43   #10
GARYP166
 
GARYP166's Avatar
 
Feb 2009

5·7 Posts
Default

Quote:
Originally Posted by Mini-Geek View Post
Are the two numbers using the same FFT length? It says the FFT length used when you start/resume the number, e.g. "Resuming primality test of M40962983 using FFT length 2560K". Your two candidates are near an FFT cutoff.
Both are using FFT length 2560K.

Quote:
Originally Posted by Mini-Geek View Post
Do you know if your CPU has two separate cores or if it is hyperthreaded? It might be related to that.
No I'm afraid I don't know that. How would I find out please?

Gary
GARYP166 is offline   Reply With Quote
Old 2009-07-13, 17:50   #11
patrik
 
patrik's Avatar
 
"Patrik Johansson"
Aug 2002
Uppsala, Sweden

52×17 Posts
Default

Quote:
Originally Posted by GARYP166 View Post
[---]

No I'm afraid I don't know that. How would I find out please?

Gary
If you are using Windows and Intel, you can download the Intel® Processor Identification Utility. Install it, and then select the CPU Technologies tab.
patrik is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
How-to guide for running LL tests on the Amazon EC2 cloud GP2 Cloud Computing 65 2021-06-14 14:05
How-to guide for running LL tests on Google Compute Engine cloud GP2 Cloud Computing 4 2020-08-03 11:21
Is it possible to disable benchmarking while torture tests are running? ZFR Software 4 2018-02-02 20:18
running single tests fast dragonbud20 Information & Answers 12 2015-09-26 21:40
Factoring Speeds Khemikal796 Lone Mersenne Hunters 5 2005-04-26 20:28

All times are UTC. The time now is 03:40.


Wed Oct 27 03:40:47 UTC 2021 up 95 days, 22:09, 0 users, load averages: 1.10, 1.21, 1.33

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.