Even if AMD’s K8 architecture has been showered with praise in many ways it is often the integrated memory controller that gets the most. By integrating the memory controller AMD has gotten a considerably more efficient use of bandwidth and most are wondering why Intel hasn’t done the same, especially considering the success of the K8. Intel has now, through its CEO Paul Otellini, revealed why it has chosen not to integrate a memory controller into its chips. The discussion was held at IDF and according to Otellini there are several important reasons for why it has chosen not to. First of all the fact that memory standards comes and goes, which means you have to redesign the whole processor when changing standard.
Intel tested the concept with an integrated memory controller with its Timna chip, a chip that was developed for Rambus. Which we now know was a major step in the wrong direction, which in turn might explain its fear for integrating the controller.
He also brought up the fact that they are going for large quantities of cache with its architectures, which makes the memory bandwidth less important. The fact that an integrated memory controller also takes up large space inside a chip was one of the negative aspects, although he forgot to mention that cache is the sole space hog we have with a processor today.
We could go on forever discussing whether Intel made a right or wrong move, but considering what we’ve seen so far of its new architecture it seems it has managed to create an extremely fast processor without an integrated memory controller.
Source: News.com
