mersenneforum.org  

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

Reply
 
Thread Tools
Old 2009-01-04, 08:33   #23
Traveller
 
May 2006
Ann Arbor, MI, USA

816 Posts
Default FWIW, I'm running v25.7

I have a network monitoring server running on an old laptop with a P-III 500. It also has only 310 Meg of RAM. It is finishing a DoubleCheck it had before the upgrade, but will be running TFs once that completes.

While it may be a bit less efficient, having the v25 client connected to the v5 server makes management much easier. Also, I have been running v24 before anyway.
Traveller is offline   Reply With Quote
Old 2009-01-04, 11:25   #24
garo
 
garo's Avatar
 
Aug 2002
Termonfeckin, IE

1010110011002 Posts
Default

I suspect that difference in 62-bit and 64-bit factoring may have to do with the way the bits are counted in v24 and v25.
garo is offline   Reply With Quote
Old 2009-01-05, 05:24   #25
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

22·41·61 Posts
Default

Quote:
Originally Posted by Traveller View Post
While it may be a bit less efficient, having the v25 client connected to the v5 server makes management much easier. Also, I have been running v24 before anyway.
Check back this post for an explanation on how you can semi-manage the pc wihile it is still using v23 or v24.
Uncwilly is offline   Reply With Quote
Old 2009-01-05, 07:28   #26
MatWur-S530113
 
MatWur-S530113's Avatar
 
Apr 2007
Spessart/Germany

2428 Posts
Default

Hello,

Quote:
Originally Posted by TheOutcaste View Post
Thought I'd post my benchmarks for V 23.8, 24.14, and 25.8 from two different PIII systems, a PIII 700 overclocked to 785 MHz, and a PIII 500.

Seems to me that sticking with V 23.8 would be more productive, at least for LL tests. Will have to run some more tests to compare factoring speeds, as V 23.8 doesn't include trial factoring in it's benchmark.

For LL tests:
V 23.8 is significantly faster on both systems
V 25.8 is slightly faster than 24.14
For Trial Factoring:
V 24.14 and V 25.8 are very close with 24.14 having a slight edge, except for 62 bit and 64 bit, where V 25.8 is approx 1.8 times as fast at 62 bit, and 2.2 times at 64 bit.
Anybody know if this is a real difference, or just due to a different number being used in the benchmark?

Code:
Intel Pentium III or Pentium III Xeon processor
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE
L1 cache size: 16 KB
L2 cache size: 512 KB
L1 cache line size: 32 bytes
L2 cache line size: 32 bytes
TLBS: 64
RdtscTiming=1
Prime95 version 23.8    Prime95 32-bit version 24.14    Prime95 32-bit version 25.8
CPU speed: 497.88 MHz   CPU speed: 497.84 MHz           CPU speed: 497.84 MHz, 2 cores
 384K FFT: 140.205 ms.    
 448K FFT: 165.715 ms.    
 512K FFT: 184.326 ms.     512K FFT:  233.152 ms.            
 640K FFT: 250.723 ms.     640K FFT:  329.114 ms.            
 768K FFT: 297.563 ms.     768K FFT:  405.953 ms.            768K FFT:  404.805 ms.
 896K FFT: 356.686 ms.     896K FFT:  484.765 ms.            896K FFT:  485.521 ms.
1024K FFT: 397.246 ms.    1024K FFT:  545.725 ms.           1024K FFT:  545.428 ms.
1280K FFT: 515.379 ms.    1280K FFT:  688.039 ms.           1280K FFT:  686.813 ms.
1536K FFT: 621.652 ms.    1536K FFT:  839.930 ms.           1536K FFT:  837.282 ms.
1792K FFT: 745.158 ms.    1792K FFT: 1013.196 ms.           1792K FFT: 1010.255 ms.
2048K FFT: 821.435 ms.    2048K FFT: 1139.053 ms.           2048K FFT: 1134.834 ms.
                          2560K FFT: 1568.413 ms.           2560K FFT: 1563.621 ms.
                          3072K FFT: 1928.861 ms.           3072K FFT: 1922.032 ms.
                          3584K FFT: 2296.835 ms.           3584K FFT: 2285.068 ms.
                          4096K FFT: 2574.921 ms.           4096K FFT: 2571.534 ms.
                          58 bit TF:  28.246 ms.            58 bit TF:  29.060 ms.
                          59 bit TF:  28.263 ms.            59 bit TF:  29.018 ms.
                          60 bit TF:  28.305 ms.            60 bit TF:  29.216 ms.
                          61 bit TF:  28.281 ms.            61 bit TF:  29.069 ms.
                          62 bit TF:  53.224 ms.            62 bit TF:  28.995 ms.
                          63 bit TF:  53.275 ms.            63 bit TF:  52.493 ms.
                          64 bit TF: 119.297 ms.            64 bit TF:  52.560 ms.
                          65 bit TF: 121.046 ms.            65 bit TF: 120.101 ms.
                          66 bit TF: 123.223 ms.            66 bit TF: 121.847 ms.
                          67 bit TF: 123.508 ms.            67 bit TF: 123.767 ms.
 
Intel(R) Pentium(R) III processor
CPU features: RDTSC, CMOV, PREFETCH, MMX, SSE
L1 cache size: 16 KB
L2 cache size: 256 KB
L1 cache line size: 32 bytes
L2 cache line size: 32 bytes
TLBS: 64
RdtscTiming=1
Prime95 version 23.8    Prime95 32-bit version 24.14    Prime95 32-bit version 25.8
CPU speed: 785.20 MHz   CPU speed: 785.66 MHz           CPU speed: 785.10 MHz
 384K FFT: 110.267 ms.    
 448K FFT: 130.652 ms.    
 512K FFT: 147.734 ms.     512K FFT:  139.285 ms.            
 640K FFT: 197.615 ms.     640K FFT:  193.755 ms.            
 768K FFT: 233.797 ms.     768K FFT:  242.919 ms.            768K FFT:  239.502 ms.
 896K FFT: 283.061 ms.     896K FFT:  292.917 ms.            896K FFT:  298.295 ms.
1024K FFT: 314.944 ms.    1024K FFT:  334.252 ms.           1024K FFT:  337.509 ms.
1280K FFT: 421.668 ms.    1280K FFT:  444.769 ms.           1280K FFT:  440.820 ms.
1536K FFT: 501.988 ms.    1536K FFT:  533.992 ms.           1536K FFT:  536.297 ms.
1792K FFT: 620.473 ms.    1792K FFT:  642.537 ms.           1792K FFT:  643.779 ms.
2048K FFT: 692.094 ms.    2048K FFT:  714.334 ms.           2048K FFT:  711.348 ms.
                          2560K FFT:  950.812 ms.           2560K FFT:  960.670 ms.
                          3072K FFT: 1176.654 ms.           3072K FFT: 1174.291 ms.
                          3584K FFT: 1412.467 ms.           3584K FFT: 1430.823 ms.
                          4096K FFT: 1643.509 ms.           4096K FFT: 1609.154 ms.
                          58 bit TF: 19.505 ms.             58 bit TF: 20.929 ms.
                          59 bit TF: 20.424 ms.             59 bit TF: 20.570 ms.
                          60 bit TF: 19.937 ms.             60 bit TF: 20.890 ms.
                          61 bit TF: 20.221 ms.             61 bit TF: 20.680 ms.
                          62 bit TF: 38.258 ms.             62 bit TF: 20.862 ms.
                          63 bit TF: 37.546 ms.             63 bit TF: 37.835 ms.
                          64 bit TF: 84.093 ms.             64 bit TF: 37.918 ms.
                          65 bit TF: 86.145 ms.             65 bit TF: 84.570 ms.
                          66 bit TF: 86.262 ms.             66 bit TF: 86.299 ms.
                          67 bit TF: 86.579 ms.             67 bit TF: 87.507 ms.
HTH

Jerry

I think what you see is a result of the 'off-by-one bug in the factoring benchmark' George Woltman talked about in this thread. In prime95 v25.8 it is corrected. Both versions of prime should need the same time for doing the same TF, but the old version reportet a wrong benchmark.

best regards,

Matthias
MatWur-S530113 is offline   Reply With Quote
Old 2009-01-05, 11:52   #27
joblack
 
joblack's Avatar
 
Oct 2008
n00bville

23·7·13 Posts
Default

Quote:
Originally Posted by axn View Post
If you ask me, ditch the PIII altogether and get a Core 2 Duo (Seriously. The Core 2 Duo will complete in 1 week, what the PIII can do in a year.)

But as long as you're sticking with the PIII, I don't believe 25.7 has any better support for PIII compared to v24.
I don´t think that the Core 2 Duo will complete in a week. My overclocked Q6600 Quad Core (4 x 3 GhZ) needs around 35 days for one mersenne number check (4 x one thread - with four threads for one numbers less but never only 1 week).

Last fiddled with by joblack on 2009-01-05 at 11:58
joblack is offline   Reply With Quote
Old 2009-01-05, 13:22   #28
axn
 
axn's Avatar
 
Jun 2003

10100000111012 Posts
Default

Quote:
Originally Posted by joblack View Post
I don´t think that the Core 2 Duo will complete in a week. My overclocked Q6600 Quad Core (4 x 3 GhZ) needs around 35 days for one mersenne number check (4 x one thread - with four threads for one numbers less but never only 1 week).
Sure. Make that two weeks. Here are the actual numbers. For the 1280K FFT, the benchmark page gives 458.11ms for P-III 600, while it gives 34ms for a Core2 Duo 2GHz. That means a productivity increase of 458/34*2 ~= 26. So what the PIII does in 52 weeks will be done by the Core 2 Duo in 2 weeks.

Of course, if we take the fastest non-OC'ed Core 2 Duo (@3.33GHz), which gives a time of 22ms, and scale down the PIII's performance by a factor of 6/5 (since OP's m/c was 500 MHz), then we get a speedup factor of 48.

No matter how you slice it, it's a no-brainer. But I have since then reconsidered my position. Go with a Core 2 Quad
axn is offline   Reply With Quote
Old 2009-01-05, 13:58   #29
joblack
 
joblack's Avatar
 
Oct 2008
n00bville

13308 Posts
Default

Quote:
Originally Posted by axn View Post
Sure. Make that two weeks. Here are the actual numbers. For the 1280K FFT, the benchmark page gives 458.11ms for P-III 600, while it gives 34ms for a Core2 Duo 2GHz. That means a productivity increase of 458/34*2 ~= 26. So what the PIII does in 52 weeks will be done by the Core 2 Duo in 2 weeks.

Of course, if we take the fastest non-OC'ed Core 2 Duo (@3.33GHz), which gives a time of 22ms, and scale down the PIII's performance by a factor of 6/5 (since OP's m/c was 500 MHz), then we get a speedup factor of 48.

No matter how you slice it, it's a no-brainer. But I have since then reconsidered my position. Go with a Core 2 Quad
Yeah the Q6600 G0 stepping is a great cpu for overclocking ... works like a charm with prime95 and is cheap for a quad core (around 150 Euro). Get a good board, some memory and a harddisk and you´re on ... all under 300 Euro.
joblack is offline   Reply With Quote
Old 2009-01-05, 22:24   #30
ixfd64
Bemusing Prompter
 
ixfd64's Avatar
 
"Danny"
Dec 2002
California

45538 Posts
Default

Versions 24.6 and later will have significantly worse performance on Pentium III processors. I think George should use some of the code from version 23.9 in future versions of Prime95, which should use such code instead of the post-23.9 code if it detects a Pentium III processor.
ixfd64 is offline   Reply With Quote
Old 2009-01-06, 04:09   #31
cheesehead
 
cheesehead's Avatar
 
"Richard B. Woods"
Aug 2002
Wisconsin USA

22·3·641 Posts
Default

IIRC, a reason for dropping optimizations for early CPU models in later prime95 versions is to keep prime95's size from growing too large to fit on a floppy.
cheesehead is offline   Reply With Quote
Old 2009-01-06, 05:04   #32
jrk
 
jrk's Avatar
 
May 2008

3·5·73 Posts
Default

People use floppies?
jrk is offline   Reply With Quote
Old 2009-01-06, 06:09   #33
joblack
 
joblack's Avatar
 
Oct 2008
n00bville

23×7×13 Posts
Default

Quote:
Originally Posted by jrk View Post
People use floppies?
In my opinion it´s not good to remove older cpu optimization even when they are only used for TF.
joblack is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
LLR Version 3.8.5 is available! Jean Penné Software 11 2011-02-20 18:22
LLR Version 3.8.0 is now available! Jean Penné Software 22 2010-04-28 07:45
LLR - new version Cruelty Riesel Prime Search 8 2006-05-16 15:00
Version 24.14 Prime95 Software 13 2006-02-15 16:32
Which LLR version to use... Cruelty Riesel Prime Search 1 2005-11-10 15:17

All times are UTC. The time now is 05:16.


Wed Oct 20 05:16:17 UTC 2021 up 88 days, 23:45, 0 users, load averages: 1.21, 1.23, 1.22

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.