Ars Technica heeft het voorlopig laatste deel van hun DRAM guide online gezet, waarin zij wat dieper in gaan op de voor- en nadelen van Rambus geheugens. De besproken onderwerpen zijn de toenemende latencies bij plaatsing van meerdere RIMMs, het Rambus Signaling Layer, het Rambus packet protocol en de power management maatregelen die zijn genomen om het stroomverbruik en de warmteontwikkeling van de Rambus chippies tot een aanvaardbaar niveau te reduceren. In de conclusie worden DDR SDRAM en RDRAM met elkaar vergeleken:
RAMBUS and DDR DRAM are two, very different next-generation memory technologies. From the ground up, the two technologies have been designed with different, yet somewhat overlapping, goals in mind. While both share the goal of providing a high bandwidth memory solution for next generation media applications, each one brings with it a unique set of advantages and disadvantages that I think this series has laid out in detail.
DDR SDRAM is an evolutionary step forward that greatly increases SDRAM technology's available bandwidth without increasing costs significantly. DDR has near price parity with the current generation of PC133 SDRAM, and it's an open standard that enjoys widespread industry support from memory makers. However, DDR SDRAM has inherited SDRAM's granularity and pin count problems--problems which will only grow worse as transistor counts increase. It's simply unavoidable.
On the other hand, RDRAM's "single chip, low pin count, high capacity, high bandwidth" solution may have sounded great in the early 90's, but the realities of today's memory market are such that nobody really cares about the "single chip, low pin count" part--at least not at this point in time in the mainstream, server, workstation, and desktop markets. And the "single chip, low pin count" part of the equation becomes even less relevant to the mainstream PC user when RAMBUS' steep price premiums are factored in.
