Originally Posted by Hellgy
This is foreseeable, How ever running a divider will not produce any better performance at the tightest timings then a ram that running on 1:1 ratioin with a CPU will on loosen timings.
Here are some results of latency Vs bandwidth: http://www.overclock.net/amd-memory/...ency+bandwidth
As you can see there's no generic answer to what is best. It's about what application you're using and getting the balancing act right.
Eg. 3Dmark05 score was higher with the slower memory speed and tighter timings, the SuperPi run benefitted from the increased bandwidth.