mersenneforum.org  

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

Reply
 
Thread Tools
Old 2007-03-15, 17:43   #1
Flatlander
I quite division it
 
Flatlander's Avatar
 
"Chris"
Feb 2005
England

31×67 Posts
Default Best Core2Duo specifically for LLR.

As I am about to build a C2D system this post got me thinking:
http://mersenneforum.org/showpost.ph...7&postcount=18

Does anyone have any statistics showing where this performance hit happens with ks of say, 16383, 32767, 65535?

Would I get better performance from a 4mb L2 cache C2D even though it might not overclock as high?
Flatlander is offline   Reply With Quote
Old 2007-03-15, 20:56   #2
juhok
 
juhok's Avatar
 
Mar 2007

2·3 Posts
Cool

This might not fully answer your question, but here's a comparison I did between my ex E6400(2MB cache) @3.2G and new E6600 @3.2G (4MB cache). E6600 falsely reported as 3.6GHz. Both were running same RAM at 400MHz(1600FSB) with same timings and same motherboard.

What I make of these results is that cache makes little to no difference at all.

Code:
Intel(R) Core(TM)2 CPU          6400  @ 2.13GHz
CPU speed: 3200.01 MHz
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
Prime95 32-bit version 24.14, RdtscTiming=1
Best time for 512K FFT length: 8.124 ms.
Best time for 640K FFT length: 10.814 ms.
Best time for 768K FFT length: 13.272 ms.
Best time for 896K FFT length: 15.721 ms.
Best time for 1024K FFT length: 18.199 ms.
Best time for 1280K FFT length: 22.302 ms.
Best time for 1536K FFT length: 27.213 ms.
Best time for 1792K FFT length: 32.395 ms.
Best time for 2048K FFT length: 35.979 ms.
Best time for 2560K FFT length: 47.562 ms.
Best time for 3072K FFT length: 57.780 ms.
Best time for 3584K FFT length: 69.326 ms.
Best time for 4096K FFT length: 77.293 ms.
Best time for 58 bit trial factors: 3.517 ms.
Best time for 59 bit trial factors: 3.520 ms.
Best time for 60 bit trial factors: 3.503 ms.
Best time for 61 bit trial factors: 3.490 ms.
Best time for 62 bit trial factors: 5.611 ms.
Best time for 63 bit trial factors: 5.592 ms.
Best time for 64 bit trial factors: 5.341 ms.
Best time for 65 bit trial factors: 5.326 ms.
Best time for 66 bit trial factors: 5.328 ms.
Best time for 67 bit trial factors: 5.295 ms.
Code:
Intel(R) Core(TM)2 CPU          6600  @ 2.40GHz
CPU speed: 3599.98 MHz
CPU features: RDTSC, CMOV, Prefetch, MMX, SSE, SSE2
L1 cache size: 32 KB
L2 cache size: unknown
L1 cache line size: 64 bytes
L2 cache line size: unknown
Prime95 32-bit version 24.14, RdtscTiming=1
Best time for 512K FFT length: 7.948 ms.
Best time for 640K FFT length: 10.768 ms.
Best time for 768K FFT length: 13.249 ms.
Best time for 896K FFT length: 15.855 ms.
Best time for 1024K FFT length: 17.543 ms.
Best time for 1280K FFT length: 22.340 ms.
Best time for 1536K FFT length: 27.247 ms.
Best time for 1792K FFT length: 32.355 ms.
Best time for 2048K FFT length: 36.072 ms.
Best time for 2560K FFT length: 47.538 ms.
Best time for 3072K FFT length: 57.882 ms.
Best time for 3584K FFT length: 70.023 ms.
Best time for 4096K FFT length: 78.621 ms.
Best time for 58 bit trial factors: 3.500 ms.
Best time for 59 bit trial factors: 3.535 ms.
Best time for 60 bit trial factors: 3.474 ms.
Best time for 61 bit trial factors: 3.492 ms.
Best time for 62 bit trial factors: 5.600 ms.
Best time for 63 bit trial factors: 5.577 ms.
Best time for 64 bit trial factors: 5.343 ms.
Best time for 65 bit trial factors: 5.323 ms.
Best time for 66 bit trial factors: 5.321 ms.
Best time for 67 bit trial factors: 5.284 ms.
juhok is offline   Reply With Quote
Old 2007-03-16, 09:51   #3
S485122
 
S485122's Avatar
 
Sep 2006
Brussels, Belgium

33×61 Posts
Default

Flatlander the post you refer to speaks about the quadcore, the QX6700 not dual cores.

Then prime95 does not properly recognises the cache of the core2 processors. It is difficult because it is shared cache : the 6400 has 2MB L2 cache for BOTH cores and the 6600 has 4MB for BOTH cores. This means that the cited benchmarks by juhok are not totally relevant. One should set the cache size in local.ini or locaxxxx.ini if running more than one instance.

Last thing is that the benchmark is not the best tool to measure overall Prime95 performance of a system. It is better to test with some real work via the advanced menu. This is especially true when running more than one instance (as one would on a dual core.)

To try to answer your original question : more cache would indeed give more speed as would a higher processor speed. One would need more data (how far can the 6400 and the 6600 be overclocked, the timings of real tests with both cores busy...) to have a definite answer.
S485122 is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Tesla cards (specifically C1060) MacFactor GPU Computing 17 2015-02-16 07:57
Prime95 for core2duo under Win32? ewmayer Software 8 2008-01-25 17:48
Sticky specifically for sieving? jasong Conjectures 'R Us 2 2008-01-16 23:31
Core2Duo overclocking report garo Hardware 24 2007-02-09 17:10

All times are UTC. The time now is 12:52.

Thu Jan 28 12:52:22 UTC 2021 up 56 days, 9:03, 0 users, load averages: 2.07, 1.90, 1.82

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.