mersenneforum.org  

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

Reply
 
Thread Tools
Old 2004-07-10, 15:05   #1
optim
 
optim's Avatar
 
Nov 2003
European Union

23·13 Posts
Default nForce2 vs. KT600

Can someone tell me why the difference in MPrime benchmark 1792K FFT between these two machines? is it the amount of memory, the dual channel, of the Nvidia/VIA chipsets?

Machine1:
Athlon 2600+ Barton
Asus systemboard with NVidia nForce2 chipset
1GB Kingston ValieRAM DDR333 RAM, 2x512MB, dual channel
SuSE Linux 9.1 OS
150ms in 1792K FFT MPrime benchmark

Machine2:
Athlon 2600+ Barton
Abit systemboard with VIA KT600 chipset
512MB Kingston ValieRAM DDR333 RAM, 1x512MB, single channel
SuSE Linux 9.1 OS
177ms in 1792K FFT MPrime benchmark
optim is offline   Reply With Quote
Old 2004-07-11, 14:45   #2
Jeff Gilchrist
 
Jeff Gilchrist's Avatar
 
Jun 2003
Ottawa, Canada

2·32·5·13 Posts
Default

Is probably a combination of the Dual channel RAM (since mprime does push a lot of data around memory) and also possibly differences in the Nvidia vs VIA chipsets themselves.

Try pulling 1 stick of RAM chips from your NForce board so it goes back to single channel mode and re-bench to see the difference.

Jeff.
Jeff Gilchrist is offline   Reply With Quote
Old 2004-07-11, 23:02   #3
E_tron
 
E_tron's Avatar
 
Sep 2002
Austin, TX

3·11·17 Posts
Default

it's the dual channel memory. 2.66Gb/sec vs. 5.4Gb/sec... yeah, its the memory.

Here is another example, my older K7 Athlon and Thunderbird Athlon machines. Their motherboards will take DDR or SDR memory(100mhz FSB). 1.6Gb/sec vs. 790Mb/sec. On average, they will lose about 30ms with SDR memory.
E_tron is offline   Reply With Quote
Old 2004-07-12, 00:38   #4
optim
 
optim's Avatar
 
Nov 2003
European Union

6816 Posts
Default

what matters most in MPrime? memory bandwidth or memory latency?
optim is offline   Reply With Quote
Old 2004-07-12, 09:00   #5
xtreme2k
 
xtreme2k's Avatar
 
Aug 2002

2·3·29 Posts
Default

To me the problem might be something else other then the memory bandwidth.

Athlon has a 64bit FSBat 400MHz it has a bandwidth of 3.2GB/s. Giving it dual channel will make HARDLY any (<5%) difference in performance. Because the CPU cannot make use of the extra bandwidth. Dual Channel DDR400 gives 6.4GB/s, while your CPU only takes 3.2GB/s. If your CPU is on 333MHz FSB it will only take 2.7GB/s.

Are you 100% sure the 2600+ are both the same. You know some AMD chips has different MHz and FSB but gives the same PR rating, that might account for the difference in speed. ANother idea might be some programs might be hogging your CPU while you do the benchmark. Or maybe a combination of things. Slower chipset and slower BIOS settings etc etc.

To me Prime95 is extremely optimised on for caches therefore the memory speed will only make a very SMALL difference. But thats on the P4. On P4 on 533FSB Bus, Single DDR333, vs a P4 on 800Mhz Bus and Dual channel DDR400, the difference per clock is EXTREMELY small. However this maybe different on the Athlon I am not too sure.

Last fiddled with by xtreme2k on 2004-07-12 at 09:03
xtreme2k is offline   Reply With Quote
Old 2004-07-12, 15:15   #6
optim
 
optim's Avatar
 
Nov 2003
European Union

23·13 Posts
Default

the CPUs are exactly the same, same MHz.
optim is offline   Reply With Quote
Old 2004-07-12, 18:47   #7
Jeff Gilchrist
 
Jeff Gilchrist's Avatar
 
Jun 2003
Ottawa, Canada

2·32·5·13 Posts
Default

Quote:
Originally Posted by optim
the CPUs are exactly the same, same MHz.
Did you try yanking the extra memory stick out of your NForce board and re-test to see the difference?

Jeff.
Jeff Gilchrist is offline   Reply With Quote
Old 2004-07-15, 05:18   #8
outlnder
 
outlnder's Avatar
 
Aug 2002

1001111102 Posts
Default

How can you possibly compare an NVidia chipset to a Via chipset.

I thought it was common knowledge that Via chipsets suck.
outlnder is offline   Reply With Quote
Old 2004-07-19, 18:19   #9
E_tron
 
E_tron's Avatar
 
Sep 2002
Austin, TX

10618 Posts
Default

Quote:
Originally Posted by outlnder
How can you possibly compare an NVidia chipset to a Via chipset.

I thought it was common knowledge that Via chipsets suck.
no... it's just nVidia is far superior to its competitors
E_tron is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
HOWTO: enable the HLT command on nforce2 chipsets with freeware utils! graysky Hardware 4 2004-10-17 10:23

All times are UTC. The time now is 07:10.

Mon Mar 8 07:10:03 UTC 2021 up 95 days, 3:21, 0 users, load averages: 2.16, 2.35, 2.38

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.