RAM - Memory Technology Overview

Introduction

If you have ever asked for advice on what type of RAM to purchase, you were probably given a response that goes something as follows. The higher the supported bandwidth of the RAM, the better, but having lower timings is also important. Both options of course bring along price increases, and buying cheaper "generic" RAM is seldom recommended. The old adage of "you get what you pay for" is often used. This is all sound advice, but what does it really mean? Why is CL2 RAM better than CL2.5 or CL3? What about the other features, like bandwidth? Higher bandwidth is often mutually exclusive to having the best timings, so which is better?

We have planned a series of articles to cover this topic, and in this first part, we will provide a general overview of memory technology and terms, and how they actually affect performance. Specific examples using one type of RAM may be included, but we are saving a detailed comparison of the DRAM types for the next article in this series. For this introduction, we are interested in growing the base of knowledge upon which future articles can be built.

Before we get into the details of memory designs, we would like to simply state up front that anyone who thinks that DRAM technology, design, and implementation are "easy" or "boring" relative to the higher profile CPU and GPU chips is guilty of a gross generalization. While it might be true that an individual DRAM chip is less complex, the memory subsystem as a whole can easily rival other parts for complexity. It should come as little surprise that in a modern PC, system RAM can often cost more on its own than any other component, with the exception of bleeding edge processors and graphics cards. The design problems and solutions can be every bit as interesting - or difficult, if you prefer - as pipelines, execution units, and cache sizes.