memory divisors versus multipliers

[VirtualLarry]

Did you ever happen to notice, that on current Intel 775-supporting chipset boards, the memory ratios are all multipliers, with the lowest being a 1:1 setting with the true FSB?

But on an AMD S939-supporting chipset board, the memory ratios all seem like divisors, with the highest setting being 1:1 with the FSB?

I'm just curious why this is so.

(And for a third data-point, NVidia 650i and 680i chipsets allow a totally async memory and FSB ratio, with individual frequencies for each.)

I for one, would like it if I could run slower memory speeds with a higher CPU FSB on C2D rigs, but it seems as though this simply isn't possible on Intel chipsets. Not so with AMD64, for those systems it's trivial to get a higher CPU FSB while still using slower RAM with a divisor.

It seems mostly pointless to run RAM at a faster speed than the processor's FSB anyways, although I realize that peripherals need RAM bandwidth too (AGP/PCI-E, etc.)

I should note that there is an exception, I run a C2D rig with an intel 865PE chipset, which does have memory divisors and not multipliers.

 

[imported_Gauss]

The divisor 4:3 is available in 945P/975X chipset motherboards in the 1066 strap.

 

[Peter]

Simple answers: The architectures are different.

AMD has the RAM controller on the CPU, running at the processor's core (!) frequency. RAM bus frequency is divided down from there. E.g. a 2400 MHz AMD64 processor would run DDR400 memory on CPU/12 = 200 MHz.

Intel's architecture has CPU and RAM connect on the northbridge, which runs both busses either independently clocked or running at a certain ratio. What the ratio choices are depends on the particular chipset.