Running Dual Channel Decreases My Memory Performance

[marks70]

I have an NF7 w/ a Barton 2500+ @ 200x11. I've been using a single 512MB of Corsair XMS PC3200 RAM (CMX512-3200C2) for the last year and just yesterday installed a second one in a dual channel configuration (slots 2 & 3). I left the timings alone at 6-3-3-2.

When I ran PCMark2002, my memory scored dropped from 6200 to 5900. Why would this be? I expected my score to go up. Do I need to change the timings? I tried 7-3-3-2.5 and 11-3-3-2.5, and that didn't do squat. Any ideas?

 

[Peter]

Socket A processors can't even use the extra bandwidth. This is no news - the CPU bus is 200 MHz DDR, 64-bit width. There's little point in making the RAM bus 200 MHz DDR 128-bit.

NVidia did that to have the surplus bandwidth available to their integrated graphics unit - before they decided to make a marketing feature of it. The subsequent dual-channel-RAM effort from VIA (KT880) proved the same thing: It's pointless.

The only benefit is that with two sets of bus drivers, you don't have to relax your timings as you add DIMMs.

 

[ts3433]

Just leave it in single channel--Peter is correct.

 

[marks70]

So I understand why I wouldn't get an increase in performance due to your explanation, but why the decrease?

 

[marks70]

Bump! Any other input appreciated!

 

[Vette73]

Well for one that is a benchmark, not real world.

The reduction MIGHT be from a latency issue? But even though the AthlonXP does not need DC mem like Intel, it can still help. Mots revoews show gains in the low single % marks, BUT a couple areas actually make the high single to low double % gains.


Do some real world test and see what that shows. I ran my old NF7-S with dual Ch and was happy.

 

[Dethfrumbelo]

Maybe running dual channel drops your command rate to 2T.

 

[Peter]

Certainly not. Having too many DIMMs on ONE channel may necessitate relaxed timings. Didn't I /just/ say that?