Memory Ratio Tech Question

[b4u]

Hi,

Using a memory ratio (for example 5:4, 3:2), will in some way put some breaks on the flow between CPU & MEM, so as I'm thinking, the CPU would still be fastest, but some speed-break would occur. More technicaly, and just to feed my mind, exactly how would this break occur, and where?


Thanks

 

[b4u]

bump!

Any info/link/... around?


Thanks

 

[pspada]

no habla yer question.

 

[IaPuP]

If you're asking why a symmetric bus/memory ratio is faster than an asymmetric, it's simple.

Symmetric, the data just flys right through the northbirdge... Think of two conveyor belts with buckets that move at the same speed. Things can automatically be dropped onto the second one at full speed without worries that the contents one one bucket might miss the contents of the bucket on the second belt, assuming they're aligned correctly.

Asymmetric, the data must be held for a cycle or part of a cycle to make sure that it is transmitted on the clock of the other bus. Imagine two conveyor belts, where one is a bit faster than the other. If it just drops it's stuff onto the other, it will drop them at a different speed and most of the bucket's contents will miss the buckets on the second belt because they're moving at a different rate.

Because of that, there would have to be a "holding area" where the contents were held, waiting for the slower belt to catch up. If the faster belt was the second one, then the slower belt could never fill up the faster one completely anyway...

*shrugs*

Eric