Dual channel DDR better than 5-10% increase

[SimMike2]

I was fooling around with my socket 939 board and dual core opteron CPU. Originally I ran this with three sticks of 512MB memory, which obviously didn't use dual channel. I decided to try my two matched sticks and take out the third stick, thus running with less memory, 1GB instead of 1.5GB, but in dual channel mode. According to Super-PI, my performance for this did improve between 5% and 10%.

But it feels much more responsive to me as I use the computer. It just seems like it runs much smoother, with the extra little torque in increased bandwidth. It seems to use my hard drive more efficiently.

Am I just imagining this, or does going dual channel make a bigger difference than the 5-10% increase in running the benchmarks?

BTW, cas settings and memory timing was the same with either configuration.

 

[imported_Septimus]

Comparing SuperPi or any benchmark to real-world performance is tough because what 'feels' better to you depends entirey on what you do on a regular basis. If you're anything like me, you probably don't sit with your computer and calculate pi on an everyday basis.

Benchmarks are more informative of your computer's potential than they are of what you'd actually notice every day -- that's why most comphreneisve benchmark programs come in 'suites' that test a bazillion different things.

It's entireyl possible that, although the specific math behind SuperPi is 5-10% better, the combination of your daily tasks coupled with the placebo effect that 'you know it's better' is producing real-world gains of 15-20%, plus or minus a few for your imagination, which usually helps...but they don't have a benchmark for that.