Random Access Memory
A computer has two kinds of memory. The first is called primary storage, and is used to hold program instructions and data while the computer is running. The other kind, called secondary storage, is used for storage of data even while the computer is not running. The primary storage is memory of a kind called Random Access Memory, RAM for short. It is volatile, i.e., susceptible to loss easily upon power outage. The secondary storage can be of many kinds--hard drives, disk drives, tape drives, and so on. The secondary storage is persistent, i.e., retains its contents even when the power is switched off.
RAM is said to be "random access" because any memory cell in RAM can be accessed directly, if the physical address of that cell is known. The counterpart of RAM is SAM, Sequential Access Memory. This kind of memory requires sequential access, i.e., access to items in the order in which they are stored. One example of a sequential-access device from common experience is a cassette tape and its recorder/player--if one wishes to listen to the fifth song on a music cassette, for instance, one has no choice but to go past the first four (either by playing or fast-forwarding). A random-access device from common experience is a music CD player, which by contrast allows for the fifth song on a CD to be played directly.
For obvious reasons, random access is preferable to sequential access—witness the fact that CD players have largely supplanted cassette players in the music business. However, there are some cases where SAM works well enough to be used--such as in memory buffers, where data are normally accessed only in the order in which they were written. One example of such a SAM buffer is the texture buffer memory in a video card.
RAM is implemented in a manner very similar to microprocessors, using an integrated circuit consisting of transistors and capacitors. The common form of RAM device is called a Dynamic Random Access Memory (DRAM), and uses a pair consisting of a transistor and a capacitor, to create a cell that can store a single bit of data. The function of the capacitor is to maintain an electric potential difference as appropriate to store a 0 or a 1; the transistor acts as a tiny switch to allow read and write access to the capacitor. Usually, the default value of a capacitor when there is no potential difference is considered a 0, while an appropriate potential counts as a 1.
There is a significant problem to be faced—the capacitor in a DRAM memory cell cannot hold its state for long. Within a few thousandths of a second, any potential difference that may have been imposed on it is lost due to leakage, thus erasing any bit information that may have been stored. Thus, for DRAMs to work, it is necessary for either the CPU or a dedicated memory controller device to come along every so often and refresh the values stored in the memory cells. This refresh must be done by first reading the value that exists in each cell, and then writing it right back on to it, thus prolonging its life. This process has to carry on automatically thousands of times a second--hence the name, Dynamic RAM. It takes a very small amount of time, usually a few dozen nanoseconds (billionths of a second) to read and recharge a memory cell.
DRAM memory cells are created by a Very-Large-Scale Integration (VLSI) etching process in an array, whose columns are called bitlines, and whose rows are called wordlines. The address of an individual memory cell is completely specified by its bitline and wordline. In order to access the memory cells, a set of specialized circuitry exists as part of the DRAM device to perform support functions.
The other major kind of RAM is static RAM (SRAM), which does not need refreshing. Bits are stored in memory using logic gates known as flip-flops, with one flip-flop required to store each bit. To implement a flip-flop requires about a half-dozen transistors along with appropriate wiring. Because no time is taken for refreshing, static RAM is significantly faster than DRAM. However, because more hardware is required per bit of storage, static RAM is more expensive for a given storage size, as well.
In CPU caches, speed is of the essence—a 1 gigahertz CPU has a cycle time of one nanosecond, and waiting for a DRAM to be read in 50 or so nanoseconds is very inefficient because many cycles have to be wasted in the wait. The size of the cache is also quite small (rarely exceeding a few hundred kilobytes, and often significantly less). For these reasons, static RAM is preferred in CPU caches. Dynamic RAM is used for the computer's main primary storage, where the speed requirement is less severe but storage has to be a lot larger.
This is the complete article, containing 812 words
(approx. 3 pages at 300 words per page).
