mersenneforum.org  

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

Reply
 
Thread Tools
Old 2013-06-22, 11:03   #1
MacMagnus
 
Jun 2013

2×3 Posts
Default The best work for my CPU

As I understand it, the "calculation-code" in Prime95 are written in Assembly-language. Doesn't this mean that you have to write specific code for each CPU there is?

I'm also under the impression that some CPUs are better for calculation A, some are better for B - relative to their calculation speed, of course.

So I wonder what work is best fit for my own CPU. I have a i5-2500, not over-clocked...
MacMagnus is offline   Reply With Quote
Old 2013-06-22, 11:55   #2
Mini-Geek
Account Deleted
 
Mini-Geek's Avatar
 
"Tim Sorbera"
Aug 2006
San Antonio, TX USA

17·251 Posts
Default

I'm pretty sure the assembly language involved here is x86. This is not CPU-specific, and covers both Intel and AMD CPUs. While Prime95 is updated as CPUs come out to take advantage of new features (e.g. SSE2, AVX), I think the core code runs on every x86/x86-64 CPU.

For the practical results of how Prime95 runs on different CPUs, see mersenne.ca's CPU throughput calculator. E.g. this comparison of your CPU and an AMD CPU is an example of how Intel CPUs run FFTs (LL, P-1, etc.) better, and AMD CPUs run TF better (relatively).

Your i5-2500 is probably best suited to FFTs - P-1, DC, or LL. For GIMPS, that's all I'd run on a CPU, since GPUs are so much better at TFing.

Last fiddled with by Mini-Geek on 2013-06-22 at 12:00
Mini-Geek is offline   Reply With Quote
Old 2013-06-22, 12:09   #3
Unregistered
 

2·571 Posts
Default

Quote:
Originally Posted by Mini-Geek View Post
I'm pretty sure the assembly language involved here is x86. This is not CPU-specific, and covers both Intel and AMD CPUs. While Prime95 is updated as CPUs come out to take advantage of new features (e.g. SSE2, AVX), I think the core code runs on every x86/x86-64 CPU.

For the practical results of how Prime95 runs on different CPUs, see mersenne.ca's CPU throughput calculator. E.g. this comparison of your CPU and an AMD CPU is an example of how Intel CPUs run FFTs (LL, P-1, etc.) better, and AMD CPUs run TF better (relatively).

Your i5-2500 is probably best suited to FFTs - P-1, DC, or LL. For GIMPS, that's all I'd run on a CPU, since GPUs are so much better at TFing.
Thanks!

Yeah I think I read somewhere that GPUs have become much better than CPUs for TF. So, then I think it is weird that the "automatically" assignment from GIMPS (Prime95) just gave me TF assignments.
  Reply With Quote
Old 2013-06-22, 13:01   #4
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

2·47·101 Posts
Default

Quote:
Originally Posted by Unregistered View Post
Thanks!

Yeah I think I read somewhere that GPUs have become much better than CPUs for TF. So, then I think it is weird that the "automatically" assignment from GIMPS (Prime95) just gave me TF assignments.
How many hours a day did you tell the program that it was likely to run? Also, are you letting it run on all cores or only 1?
Uncwilly is online now   Reply With Quote
Old 2013-06-22, 13:52   #5
MacMagnus
 
Jun 2013

1102 Posts
Default

Quote:
Originally Posted by Uncwilly View Post
How many hours a day did you tell the program that it was likely to run? Also, are you letting it run on all cores or only 1?
24 h/day. It will run on 4/4 cores. :-)

It's a computer that I don't use for anything else.
MacMagnus is offline   Reply With Quote
Old 2013-06-22, 14:59   #6
davieddy
 
davieddy's Avatar
 
"Lucan"
Dec 2006
England

2×3×13×83 Posts
Default

Quote:
Originally Posted by Mini-Geek View Post
Your i5-2500 is probably best suited to FFTs - P-1, DC, or LL. For GIMPS, that's all I'd run on a CPU, since GPUs are so much better at TFing.
Quote:
Originally Posted by MacMagnus View Post
24 h/day. It will run on 4/4 cores. :-)
It's a computer that I don't use for anything else.
First time LLs are definitely your game.
(P-1 as necessary, and leave DCs to the slower guys).

If there is a decent GPU with your i5, join in the effort to ensure that every exponent >64M is TFed to 74. (Chalsall needs help!).

Re "Winning the lottery", 1/115000 every 3 weeks is certainly better than the norm for such things.
This applies to expos ~60M.
For a 100M digit prime, reduce your expectations by a factor of 200.
And take out some life insurance in case you are not around in 2030 AD

David
davieddy is offline   Reply With Quote
Old 2013-06-22, 15:31   #7
davieddy
 
davieddy's Avatar
 
"Lucan"
Dec 2006
England

647410 Posts
Default

Quote:
Originally Posted by Mini-Geek View Post
I'm pretty sure the assembly language involved here is x86. This is not CPU-specific, and covers both Intel and AMD CPUs.
I'm damn certain it is.
But the "non CPU specific" bit applies more to the history of hardware development than the "level" of programming.
There are things called "compilers" which occasionally ceate/modify assembler that runs on varying platforms that runs ~1/3 times as fast as I could manage (in my assembly programming prime)

D
davieddy is offline   Reply With Quote
Old 2013-06-22, 17:53   #8
TheMawn
 
TheMawn's Avatar
 
May 2013
East. Always East.

6BF16 Posts
Default

Someone please correct me if I am wrong but is it possible that the server may assign TF just to test the CPU's reliability before giving it a LL test?

Or, it is also possible that this may be one of the symptoms of the iffy server. A few people have been noting problems of late.
TheMawn is offline   Reply With Quote
Old 2013-06-22, 19:45   #9
MacMagnus
 
Jun 2013

2·3 Posts
Default

Can it be because of my BIOS-setup? I sat it to "fail safe" instead of "optimal". Sounds weird if it could be this, I didn't thing the difference were that large...
MacMagnus is offline   Reply With Quote
Old 2013-06-22, 20:06   #10
MacMagnus
 
Jun 2013

2×3 Posts
Default

Quote:
Originally Posted by davieddy View Post
First time LLs are definitely your game.
(P-1 as necessary, and leave DCs to the slower guys).

If there is a decent GPU with your i5, join in the effort to ensure that every exponent >64M is TFed to 74. (Chalsall needs help!).

Re "Winning the lottery", 1/115000 every 3 weeks is certainly better than the norm for such things.
This applies to expos ~60M.
For a 100M digit prime, reduce your expectations by a factor of 200.
And take out some life insurance in case you are not around in 2030 AD

David
Alas I don't even have a separate GPU. (I'm not a gamer! )

A reduce by a factor of 200? Some lotteries have those odds, hehe.

No, I don't expect to "win". It's fun tho, doing scientific research, hehe.
MacMagnus is offline   Reply With Quote
Old 2013-06-22, 23:19   #11
davieddy
 
davieddy's Avatar
 
"Lucan"
Dec 2006
England

11001010010102 Posts
Default

Quote:
Originally Posted by MacMagnus View Post
Alas I don't even have a separate GPU. (I'm not a gamer! )

A reduce by a factor of 200? Some lotteries have those odds, hehe.

No, I don't expect to "win". It's fun tho, doing scientific research, hehe.
Sounds like you'll fit in fine here.
Have you been "Lurking"?

D
davieddy is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
How to calculate work/effort for PRP work? James Heinrich PrimeNet 0 2011-06-28 19:29
No Work Pilgrim Information & Answers 1 2008-01-31 18:53
Out of Work? birdman2584 Sierpinski/Riesel Base 5 12 2006-11-22 00:06
It seems to work, but why ? T.Rex Math 15 2005-10-15 10:38
work to do... guido72 Software 2 2002-09-26 15:47

All times are UTC. The time now is 14:10.

Tue Apr 20 14:10:16 UTC 2021 up 12 days, 8:51, 0 users, load averages: 3.73, 3.91, 3.76

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.