That would only be true if efficency was 100%, which never happens. I've been able to get 96%+ efficency (in buffered reads), but only with high clocked S754 AMD chips that have a single memory channel and on-die memory controlers, that were paired with Winbond memory at rediculously tight timings.
I see quite a significant difference in memory performance on the system in my sig going from DDR-800 3-3-3-8 to DDR-1200 5-5-5-12, both at the same 400MHz FSB, with the same tRD fo 5 (the lowest possible at either setting). The DDR-1200 CL5 settings are nearly 15% faster in both read bandwidth and latency. Of course real world performance differences are less, as memory is only one part of a much larger equation, I can still see decent difference in benchmarks, and even feel a modest improvement in responsiveness (both in windows, and in games).
DDR3 with it's reduction in efficency compared to DDR2, needs higher clocks (which it has) at the same FSB to compensate.
Even NVIDIA chipsets efficency gains at 1:1 are unlikly to overcome the massive deficit in clock rate with DDR-3.
Originally Posted by Brenbren
so instead of pc2 6400 ram for a 400mhz bus speed, you would want pc3 16000? to get 1:2 instead of 1:1?
PC3-12800/DDR3-1600 is a fairly good match for a 400MHz FSB.