Memory

 From the point of view of a software developer, the memory system is the most visible part of the computer. If we didn’t have memory, we’d never have a problem with an errant pointer. But that’s another story. The computer memory is the place where program code (instructions) and variables (data) are stored. We can make a simple analogy about instructions and data. Consider a recipe to bake a cake. The recipe itself is the collection of instructions that tell us how to create the cake. The data represents the ingredients we need that the instructions manipulate. It doesn’t make much sense to sift the flour if you don’t have flour to sift.We may also describe memory as a hierarchy, based upon speed. In this case, speed means how fast the data can be retrieved from the memory when the computer requests it. The fastest memory is also the most expensive so as the memory access times become slower, the cost per bit decreas-es, so we can have more of it. The fastest memory is also the memory that’s closest to the CPU. Thus, our CPU might have a small number of on-chip data registers, or storage locations, several thousand locations of off-chip cache memory, several million locations of main memory and several billion locations of disk storage. The ratio of the access time of the fastest on-chip memory to the slowest memory, the hard disk, is about 10,000 to one. The ratio of the cost of the two memories is somewhat more difficult to calculate because the fastest semiconductor memory is the on-chip cache memory, and you cannot buy that separately from the microprocessor itself. How-ever, if we estimate the ratio of the cost per gigabyte of the main memory in your PC to the cost per gigabyte of hard disk storage (and taking into account the mail-in rebates) then we find that the faster semiconductor storage with an average access time of 20–40 nanoseconds is 300 times more costly then hard disk storage, with an average access time of 1 millisecond. Today, because of the economy of scale provided by the PC industry, memory is incredibly inexpensive. A standard memory module (SIMM) with a capacity of 512 million storage locations costs about $60. PC memory is dominated by a memory technology called dynamic random access memory, or DRAM. There are several variations of DRAM, and we’ll cover them in greater depth later on. DRAM is characterized by the fact that it must be constantly accessed or it will lose its stored data. This forces us to create highly specialized and complex support hardware to interface the memory systems to the CPU. These devices are contained in support chipsets that have be-come as important to the modern PC as the CPU. Why use these complex memories? DRAM’s are inherently very dense and can hold upwards of 512 million bits of information. In order to achieve these densities, the complexity of accessing and controlling them was moved to the chipset