Raw compute performance difference between M4400 and M6400

The M4400 supports dual-channel DDR2-800 memory.
The M6400 supports dual-channel DDR3-1066 memory.
Both systems have a FSB that operates at 1066 MHz.

I'm trying to understand if the different memory technologies ultimately make a difference for computational (i.e. math on large data sets) performance in these particular machines.

Consider an M4400 and M6400 with identical CPUs and amounts of DDR-2 and DDR-3 memory, respectively. And suppose that all relevant memory latencies are identical. For the sake of this question, I don't care about power consumption and heat (which is obviously a benefit of DDR3).

Question:
Since both machines support dual-channel memory operation, the maximum effective bandwidth of the memory bus is doubled. So the dual-channel DDR2-800 can achieve 12.8 GB/sec transfer rate, while the dual-channel DDR3-1066 can achieve 17 GB/sec transfer rate. In either case, these rates are higher than the 8.5 GB/sec rate that the FSB can achieve, so it seems that performance in both machines is limited by the speed of the FSB (1066 MHz). Is that correct?

For more info on matched FSB and memory ratings under Core 2, see:
http://en.wikipedia.org/wiki/Intel_Core_2#Synchronous_memory_modules

 

So far tests have indicated a very minimal performance gain with DDR3 memory.

If you don't need the video card or more than 8gb of ram, I would go for the M4400.

Greg

 

Oh man, that sucks that the m4400 does not support ddr3

 

Agreed - Other differences between them for me are the num pad and RAID 0/1. The M4400's secondary hard drive (in place of the optical drive) must be a 1.8" model rather than a 2.5" limiting its usefulness.

N

 

What about not supporting DDR3 sucks and how would it benefit you if it did support DDR3?

 

Not much actually. 4% difference in DDR2-800 and DDR3-1066.

(True/Symmetric) Dual Channel boosts bandwidth by 23%.

DDR2/DDR3 RAM guide

 

LinuxBuny,

the point you question is excellent. Congratulations.
I always search a such test/comparatif/benchmark but i didn't find out one !

2 Month aga, i ask the world :

What is the performance factor between this 2 configurations :

- CPU T9600 2.8GHz FSB 1066 with DDR3-1066 memory
- CPU X9000 2.8GHz FSB 800 with DDR2-800 memory

Both are same cpu speed but different architecture (one is centrino1, other is centrino2) with different FSB and RAM speed.

If anyone have an article, test, i will be VERY interesting !!!!!!!!!!!

ps : don't speak about heat and energy, we know centrino2 is less heat and consum less.

 

Centrino 2 provides a useful boost to the memory bandwidth. However, how this affects performance depends on what you are doing. In many activities the memory bandwidth is not a significant factor.

John

 

no diference between ddr3 and ddr2 when it comes to wprime cpu test.
been there, tested that.

 

Thanks for affirming my question. For the record, it's linux*BUM*, not BUNY! No respect...

Forget about DDR2 vs. DDR3 for the moment. The upshot of my original post is that it would seem that when using dual-channel memory (at 800 or 1066 MHz), the peak memory bandwidth is limited by the FSB (at 1066 MHz), so it shouldn't matter, *even in principle*, which memory speed (either 800 or 1066 MHz) you have.

In fact, by the same reasoning (i.e. that dual-channel mode doubles memory bandwidth), one would expect no degradation in memory bandwidth even with "slow" dual-channel 533 MHz memory!