Maybe it's for granularity (being favourable to latency also), just as GeForce3/4 have 4 independant memory-controllers working together to have the same advantageQuote:
Originally posted by elimc
One thing I'm curious about naukkis: if you don't think RAM latencies are that important, then why is the EV7 built with 8 small channels of RDRAM and not one or two large channels of RDRAM? The way the EV7 bus is designed gives me the impression that they tried to keep the latencies to a minimum.
you can either acces multiple smaller datachunks separated over the controllers, or use them together to acces one big datachunk, depending on what you need
