mersenneforum.org  

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

Reply
 
Thread Tools
Old 2008-08-13, 10:23   #1
Primix
 
Jul 2008

5 Posts
Default A program that uses all the CPU-cores

I have a dual core CPU. I'm currently running Prime95 24.14. It doesn't use more than half the CPU-capacity. I also tried Prime95 25.6, which resulted in 100% CPU usage. But to do that, it had to run two "searches" (simultaneously), so the effect of running two instances of 24.14 would be just the same.

I wonder if there in the future will be a program that make 100% use of the CPU-capacity, but only searches for one prime at a time? Now my computer uses about two months on a number (running at 50%). Will there be a program that could shorten this to one month (running at 100%)?

My question does also applies to quad core CPUs.

Or is this technically impossible?
Primix is offline   Reply With Quote
Old 2008-08-13, 11:44   #2
Mini-Geek
Account Deleted
 
Mini-Geek's Avatar
 
"Tim Sorbera"
Aug 2006
San Antonio, TX USA

17·251 Posts
Default

Quote:
Originally Posted by Primix View Post
I have a dual core CPU. I'm currently running Prime95 24.14. It doesn't use more than half the CPU-capacity. I also tried Prime95 25.6, which resulted in 100% CPU usage. But to do that, it had to run two "searches" (simultaneously), so the effect of running two instances of 24.14 would be just the same.

I wonder if there in the future will be a program that make 100% use of the CPU-capacity, but only searches for one prime at a time? Now my computer uses about two months on a number (running at 50%). Will there be a program that could shorten this to one month (running at 100%)?

My question does also applies to quad core CPUs.

Or is this technically impossible?
It is possible, and in fact 25.6 has it as an option (Test > Worker Threads, set # of threads to run to 1, affinity to any CPU, and # of CPUs to use to 2), but it results in lower efficiency, so it's best to run two to four side-by-side searches. It will give you higher throughput to run it separately, it'll just (making up numbers, just approximating) finish two every 30 days instead of one every 18 days. Of course, if you want, you could have it run on both cores anyway, just be aware that you'll finish numbers slower over a long period of time.
Mini-Geek is offline   Reply With Quote
Old 2008-08-13, 12:29   #3
Primix
 
Jul 2008

58 Posts
Default

Ok, thanks! :)

Guess I'll be running two instances of the 24.14-program then.
Primix is offline   Reply With Quote
Old 2008-08-13, 14:26   #4
mdettweiler
A Sunny Moo
 
mdettweiler's Avatar
 
Aug 2007
USA (GMT-5)

186916 Posts
Default

Quote:
Originally Posted by Primix View Post
Ok, thanks! :)

Guess I'll be running two instances of the 24.14-program then.
Or, you could run v25.6, with four threads, but with each core running its own assignment (which is, in fact, the default configuration).
mdettweiler is offline   Reply With Quote
Old 2008-08-15, 16:56   #5
ixfd64
Bemusing Prompter
 
ixfd64's Avatar
 
"Danny"
Dec 2002
California

28·32 Posts
Default

The problem is that version 25.6 is still in beta, though. Beta programs tend to be unstable, and George actually recommends against installing over version 24.14, the latest stable release.
ixfd64 is offline   Reply With Quote
Old 2008-08-15, 20:39   #6
Batalov
 
Batalov's Avatar
 
"Serge"
Mar 2008
Phi(4,2^7658614+1)/2

2×33×132 Posts
Default

Using 25.6, you absolutely can run one number on two cores (or four cores on a quad). It is a bit less efficient than running two numbers, though.

Your mileage may vary - on some systems it is significantly less efficient. This depends on a lot of things; you need to run some tests. Just start by running the "Benchmark" (in the "Options" menu) and review the numbers.
Batalov is offline   Reply With Quote
Old 2008-09-06, 20:03   #7
jasong
 
jasong's Avatar
 
"Jason Goatcher"
Mar 2005

350510 Posts
Default

I realize people may make fun of me because of the other stuff I've posted a few days ago, but has anyone tested the 2 core, 1 thread(hope that's right) option on an AMD dual-core? I'm not looking for the efficiency compared to an Intel chip, but how much efficiency is lost when compared to running 2 numbers with 1 on each core of an AMD dual-core?

I've HEARD that AMD is so much better, but the stuff they use to bench the cores in the articles I read doesn't mean anything to me. I'd be very much interested in hearing how an AMD does at this, since they "have direct cache-to-cache communication on the chip" or some such thing.
jasong is offline   Reply With Quote
Old 2008-09-06, 21:09   #8
Mini-Geek
Account Deleted
 
Mini-Geek's Avatar
 
"Tim Sorbera"
Aug 2006
San Antonio, TX USA

17×251 Posts
Default

I have just such a benchmark posted in the benchmark thread:
Quote:
Originally Posted by Mini-Geek View Post
Code:
[Sat Jun 21 09:22:51 2008]
Compare your results to other computers at http://www.mersenne.org/bench.htm
AMD Athlon(tm) 64 X2 Dual Core Processor 4800+
CPU speed: 2505.98 MHz, 2 cores
CPU features: RDTSC, CMOV, Prefetch, 3DNow!, MMX, SSE, SSE2
L1 cache size: 64 KB
L2 cache size: 512 KB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
L1 TLBS: 32
L2 TLBS: 512
Prime95 32-bit version 25.6, RdtscTiming=1
Best time for 4K FFT length: 0.093 ms.
Best time for 5K FFT length: 0.130 ms.
Best time for 6K FFT length: 0.161 ms.
Best time for 7K FFT length: 0.203 ms.
Best time for 8K FFT length: 0.223 ms.
Best time for 10K FFT length: 0.275 ms.
Best time for 12K FFT length: 0.333 ms.
Best time for 14K FFT length: 0.402 ms.
Best time for 16K FFT length: 0.440 ms.
Best time for 20K FFT length: 0.568 ms.
Best time for 24K FFT length: 0.689 ms.
Best time for 28K FFT length: 0.834 ms.
Best time for 32K FFT length: 0.910 ms.
Best time for 40K FFT length: 1.254 ms.
Best time for 48K FFT length: 1.534 ms.
Best time for 56K FFT length: 1.851 ms.
Best time for 64K FFT length: 2.049 ms.
Best time for 80K FFT length: 2.770 ms.
Best time for 96K FFT length: 3.395 ms.
Best time for 112K FFT length: 4.082 ms.
Best time for 128K FFT length: 4.528 ms.
Best time for 160K FFT length: 5.445 ms.
Best time for 192K FFT length: 6.596 ms.
Best time for 224K FFT length: 7.962 ms.
Best time for 256K FFT length: 8.839 ms.
Best time for 320K FFT length: 11.292 ms.
Best time for 384K FFT length: 13.957 ms.
Best time for 448K FFT length: 16.935 ms.
Best time for 512K FFT length: 18.901 ms.
Best time for 640K FFT length: 24.724 ms.
Best time for 768K FFT length: 29.892 ms.
Best time for 896K FFT length: 35.971 ms.
Best time for 1024K FFT length: 39.913 ms.
Best time for 1280K FFT length: 50.690 ms.
Best time for 1536K FFT length: 62.344 ms.
Best time for 1792K FFT length: 75.364 ms.
Best time for 2048K FFT length: 84.073 ms.
Best time for 2560K FFT length: 112.600 ms.
Best time for 3072K FFT length: 137.808 ms.
Best time for 3584K FFT length: 166.321 ms.
Best time for 4096K FFT length: 185.784 ms.
Best time for 5120K FFT length: 249.135 ms.
Best time for 6144K FFT length: 305.457 ms.
Best time for 7168K FFT length: 374.386 ms.
Best time for 8192K FFT length: 430.404 ms.
Best time for 10240K FFT length: 543.168 ms.
Best time for 12288K FFT length: 662.485 ms.
Best time for 14336K FFT length: 806.481 ms.
Best time for 16384K FFT length: 924.455 ms.
Best time for 20480K FFT length: 1252.055 ms.
Best time for 24576K FFT length: 1576.145 ms.
Best time for 28672K FFT length: 1955.174 ms.
[Sat Jun 21 09:28:40 2008]
Best time for 32768K FFT length: 2268.767 ms.
Timing FFTs using 2 threads.
Best time for 4K FFT length: 0.092 ms.
Best time for 5K FFT length: 0.128 ms.
Best time for 6K FFT length: 0.161 ms.
Best time for 7K FFT length: 0.203 ms.
Best time for 8K FFT length: 0.223 ms.
Best time for 10K FFT length: 0.217 ms.
Best time for 12K FFT length: 0.250 ms.
Best time for 14K FFT length: 0.290 ms.
Best time for 16K FFT length: 0.318 ms.
Best time for 20K FFT length: 0.388 ms.
Best time for 24K FFT length: 0.469 ms.
Best time for 28K FFT length: 0.549 ms.
Best time for 32K FFT length: 0.601 ms.
Best time for 40K FFT length: 1.303 ms.
Best time for 48K FFT length: 1.363 ms.
Best time for 56K FFT length: 1.584 ms.
Best time for 64K FFT length: 1.678 ms.
Best time for 80K FFT length: 1.860 ms.
Best time for 96K FFT length: 2.244 ms.
Best time for 112K FFT length: 2.669 ms.
Best time for 128K FFT length: 3.033 ms.
Best time for 160K FFT length: 3.676 ms.
Best time for 192K FFT length: 4.433 ms.
Best time for 224K FFT length: 5.313 ms.
Best time for 256K FFT length: 5.976 ms.
Best time for 320K FFT length: 7.995 ms.
Best time for 384K FFT length: 9.527 ms.
Best time for 448K FFT length: 11.460 ms.
Best time for 512K FFT length: 12.866 ms.
Best time for 640K FFT length: 15.931 ms.
Best time for 768K FFT length: 19.222 ms.
Best time for 896K FFT length: 24.674 ms.
Best time for 1024K FFT length: 27.541 ms.
Best time for 1280K FFT length: 34.639 ms.
Best time for 1536K FFT length: 41.525 ms.
Best time for 1792K FFT length: 49.814 ms.
Best time for 2048K FFT length: 55.185 ms.
Best time for 2560K FFT length: 76.146 ms.
Best time for 3072K FFT length: 90.998 ms.
Best time for 3584K FFT length: 108.793 ms.
Best time for 4096K FFT length: 120.635 ms.
Best time for 5120K FFT length: 152.836 ms.
Best time for 6144K FFT length: 188.856 ms.
Best time for 7168K FFT length: 231.610 ms.
Best time for 8192K FFT length: 268.662 ms.
Best time for 10240K FFT length: 329.272 ms.
Best time for 12288K FFT length: 408.491 ms.
Best time for 14336K FFT length: 493.651 ms.
Best time for 16384K FFT length: 568.213 ms.
Best time for 20480K FFT length: 744.815 ms.
Best time for 24576K FFT length: 919.348 ms.
Best time for 28672K FFT length: 1111.659 ms.
Best time for 32768K FFT length: 1273.254 ms.
Best time for 58 bit trial factors: 4.759 ms.
Best time for 59 bit trial factors: 4.776 ms.
Best time for 60 bit trial factors: 4.754 ms.
Best time for 61 bit trial factors: 4.775 ms.
Best time for 62 bit trial factors: 8.732 ms.
Best time for 63 bit trial factors: 8.713 ms.
Best time for 64 bit trial factors: 11.035 ms.
Best time for 65 bit trial factors: 10.961 ms.
Best time for 66 bit trial factors: 10.975 ms.
Best time for 67 bit trial factors: 10.952 ms.
My average is that 2 cores takes 0.708 times the time of 1 core.
This result of Xyzzy's on a dual core Intel has an average of 0.628 times:
Quote:
Originally Posted by Xyzzy View Post
Code:
Intel(R) Core(TM)2 Duo CPU     T5750  @ 2.00GHz
CPU speed: 1994.84 MHz, 2 cores
CPU features: RDTSC, CMOV, Prefetch, MMX, SSE, SSE2
L1 cache size: 32 KB
L2 cache size: 2048 KB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
TLBS: 256
Prime95 64-bit version 25.6, RdtscTiming=1
Best time for 4K FFT length: 0.130 ms.
Best time for 5K FFT length: 0.096 ms.
Best time for 6K FFT length: 0.114 ms.
Best time for 7K FFT length: 0.140 ms.
Best time for 8K FFT length: 0.144 ms.
Best time for 10K FFT length: 0.200 ms.
Best time for 12K FFT length: 0.245 ms.
Best time for 14K FFT length: 0.297 ms.
Best time for 16K FFT length: 0.309 ms.
Best time for 20K FFT length: 0.418 ms.
Best time for 24K FFT length: 0.512 ms.
Best time for 28K FFT length: 0.627 ms.
Best time for 32K FFT length: 0.658 ms.
Best time for 40K FFT length: 0.858 ms.
Best time for 48K FFT length: 1.046 ms.
Best time for 56K FFT length: 1.278 ms.
Best time for 64K FFT length: 1.365 ms.
Best time for 80K FFT length: 1.930 ms.
Best time for 96K FFT length: 2.348 ms.
Best time for 112K FFT length: 2.828 ms.
Best time for 128K FFT length: 3.007 ms.
Best time for 160K FFT length: 3.695 ms.
Best time for 192K FFT length: 4.618 ms.
Best time for 224K FFT length: 5.562 ms.
Best time for 256K FFT length: 6.195 ms.
Best time for 320K FFT length: 8.020 ms.
Best time for 384K FFT length: 9.924 ms.
Best time for 448K FFT length: 11.862 ms.
Best time for 512K FFT length: 13.120 ms.
Best time for 640K FFT length: 17.372 ms.
Best time for 768K FFT length: 21.327 ms.
Best time for 896K FFT length: 25.340 ms.
Best time for 1024K FFT length: 29.086 ms.
Best time for 1280K FFT length: 35.711 ms.
Best time for 1536K FFT length: 45.938 ms.
Best time for 1792K FFT length: 52.387 ms.
Best time for 2048K FFT length: 58.895 ms.
Best time for 2560K FFT length: 87.575 ms.
Best time for 3072K FFT length: 111.687 ms.
Best time for 3584K FFT length: 155.181 ms.
Best time for 4096K FFT length: 168.838 ms.
Best time for 5120K FFT length: 232.465 ms.
Best time for 6144K FFT length: 300.670 ms.
Best time for 7168K FFT length: 377.410 ms.
Best time for 8192K FFT length: 423.452 ms.
Best time for 10240K FFT length: 574.246 ms.
Best time for 12288K FFT length: 658.277 ms.
Best time for 14336K FFT length: 826.424 ms.
Best time for 16384K FFT length: 924.886 ms.
Best time for 20480K FFT length: 1279.596 ms.
Best time for 24576K FFT length: 1531.288 ms.
Best time for 28672K FFT length: 1811.050 ms.
Best time for 32768K FFT length: 2088.434 ms.
Timing FFTs using 2 threads.
Best time for 4K FFT length: 0.065 ms.
Best time for 5K FFT length: 0.096 ms.
Best time for 6K FFT length: 0.115 ms.
Best time for 7K FFT length: 0.140 ms.
Best time for 8K FFT length: 0.143 ms.
Best time for 10K FFT length: 0.125 ms.
Best time for 12K FFT length: 0.147 ms.
Best time for 14K FFT length: 0.177 ms.
Best time for 16K FFT length: 0.185 ms.
Best time for 20K FFT length: 0.241 ms.
Best time for 24K FFT length: 0.292 ms.
Best time for 28K FFT length: 0.350 ms.
Best time for 32K FFT length: 0.375 ms.
Best time for 40K FFT length: 1.506 ms.
Best time for 48K FFT length: 0.994 ms.
Best time for 56K FFT length: 0.817 ms.
Best time for 64K FFT length: 1.143 ms.
Best time for 80K FFT length: 1.045 ms.
Best time for 96K FFT length: 1.261 ms.
Best time for 112K FFT length: 1.511 ms.
Best time for 128K FFT length: 1.609 ms.
Best time for 160K FFT length: 1.943 ms.
Best time for 192K FFT length: 2.418 ms.
Best time for 224K FFT length: 2.961 ms.
Best time for 256K FFT length: 3.308 ms.
Best time for 320K FFT length: 4.348 ms.
Best time for 384K FFT length: 5.579 ms.
Best time for 448K FFT length: 6.463 ms.
Best time for 512K FFT length: 10.402 ms.
Best time for 640K FFT length: 10.010 ms.
Best time for 768K FFT length: 11.841 ms.
Best time for 896K FFT length: 14.175 ms.
Best time for 1024K FFT length: 18.984 ms.
Best time for 1280K FFT length: 19.918 ms.
Best time for 1536K FFT length: 24.252 ms.
Best time for 1792K FFT length: 28.769 ms.
Best time for 2048K FFT length: 32.021 ms.
Best time for 2560K FFT length: 42.392 ms.
Best time for 3072K FFT length: 53.609 ms.
Best time for 3584K FFT length: 62.312 ms.
Best time for 4096K FFT length: 75.522 ms.
Best time for 5120K FFT length: 122.694 ms.
Best time for 6144K FFT length: 167.483 ms.
Best time for 7168K FFT length: 205.575 ms.
Best time for 8192K FFT length: 220.436 ms.
Best time for 10240K FFT length: 290.311 ms.
Best time for 12288K FFT length: 378.684 ms.
Best time for 14336K FFT length: 479.343 ms.
Best time for 16384K FFT length: 582.784 ms.
Best time for 20480K FFT length: 768.802 ms.
Best time for 24576K FFT length: 863.376 ms.
Best time for 28672K FFT length: 1117.997 ms.
Best time for 32768K FFT length: 1373.751 ms.
Best time for 58 bit trial factors: 4.499 ms.
Best time for 59 bit trial factors: 4.461 ms.
Best time for 60 bit trial factors: 4.907 ms.
Best time for 61 bit trial factors: 5.315 ms.
Best time for 62 bit trial factors: 6.059 ms.
Best time for 63 bit trial factors: 7.628 ms.
Best time for 64 bit trial factors: 8.116 ms.
Best time for 65 bit trial factors: 8.770 ms.
Best time for 66 bit trial factors: 8.730 ms.
Best time for 67 bit trial factors: 8.645 ms.
So no, AMDs aren't better for multiple. Or at least, my AMD in that test wasn't better than his Intel in that test.
Mini-Geek is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Cuda Cores storm5510 GPU Computing 32 2016-10-27 22:18
4 Cores, only 1 worker? sonjohan Software 13 2016-08-26 05:34
Best use of six cores for L-L testing Chuck Hardware 4 2011-05-10 03:04
32 cores limitation gabrieltt Software 12 2010-07-15 10:26
CPU cores Unregistered Information & Answers 7 2009-11-02 08:27

All times are UTC. The time now is 19:16.

Sun Sep 27 19:16:49 UTC 2020 up 17 days, 16:27, 0 users, load averages: 1.78, 1.86, 1.73

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.