Reference Series Table of Contents For This Issue

How Computers Work, Part I
August 2001• Vol.5 Issue 3
Page(s) 12-17 in print issue

Memories Are Made Of This
Several Types Of Memory Play A Role In PCs
In humans, itís a mysterious mental facility that allows us to remember facts and events long after we learned them or they occurred. This could be a first trip to the ballpark, the sum of two plus two, or the name of a favorite cousin. Remembrances of these and thousands of other events are retained in our minds, where theyíre accessible whenever we need to recall them. Memory is one of the human capacities that sets us apart from other living creatures. It allows us to learn and make comparisons, as well as develop and maintain relationships.

Memory, as it exists within a computer, cannot be discussed in such poetic terms. Computer memory is any area where data is stored on a semiconductor chip (a tiny board made of silicon and germanium) rather than on some other storage medium, such as a hard drive, compact disc, or tape cartridge. The data may be stored permanently or temporarily on the chip. In a computer, memory is a necessity. The more, the better.

Thatís all you really need to know about computer memory. For the most part, itís a hands-off component of the PC. The only time youíll even know itís there is when you push your PC to do too many things at once and receive an error message in response. Nevertheless, knowing how memory works is an important step in understanding how your PC manages data.

  Memory Variations. Computer memory is not a single entity, like a modem or a hard drive. Rather, itís several complementary components that work together within the PC. The component with which youíre probably the most familiar is RAM (random-access memory), which is also referred to as a computerís main memory. RAM is the area in a computer where data is held temporarily while an application or the OS (operating system) is using it. Another type of memory known as ROM (read-only memory) retains data permanently. RAM and ROM are the two types of memory.

ROM. ROM describes an area of the memory containing data that may be accessed but not altered. Many electronic devices, including calculators and digital clocks, contain ROM. In a computer, youíll usually find ROM on a memory chip that either plugs into or is soldered onto the motherboard.

Computers contain only a small amount of ROM, just enough to support the BIOS (Basic Input/Output System). The BIOS is a set of instructions that controls the computerís startup process, activates the hardware components, and loads the OS. Because the ROM chip in a computer contains only the BIOS, itís sometimes called the BIOS ROM.

In the early days of personal computing, the BIOS was permanently established in ROM by the memory manufacturer. This proved inconvenient, particularly to computer manufacturers who occasionally needed to upgrade the BIOS before selling their wares. To solve this problem, several types of PROM (programmable ROM) were developed. PROM chips are not inscribed with data during the manufacturing process. These blank chips are sold to computer manufacturers and resellers who use a special piece of hardware called a PROM programmer to burn (inscribe with data) the chips.

Several types of PROM exist, including EPROM (erasable PROM), EEPROM (electrically erasable PROM), and flash memory, which is actually a type of EEPROM. Data contained in these types of PROM can be erased and updated when needed, using a PROM programmer or a special software application.

RAM. This type of memory, which is sometimes referred to as main memory, is the largest area of chip-based memory storage in your computer system. All data contained in RAM is held there temporarily until itís no longer needed or the computer is turned off. RAM takes its name from the fact that it allows data to be accessed randomly. That means the CPU (central processing unit; the device [also called the processor] that interprets and carries out instructions in a computer) can access RAM data directly, in a non-linear fashion. Note that RAM is not the opposite of ROM. ROM is randomly accessible, too. The name RAM actually comes from its comparison with the first forms of computer memory, which stored data in a linear format.

Every PC contains at least two types of RAM: DRAM (dynamic RAM) and SRAM (static RAM). Additionally, most systems contain some form of peripheral RAM. Video memory is the most popular form of peripheral RAM, but many printers also contain peripheral RAM.

Dynamic RAM. When we talk about RAM, we usually refer to DRAM. DRAM is one of the most active components in your PC. In fact, almost everything you see on your computer screen is being held temporarily in DRAM. Many PCs sold today come equipped with 128MB of DRAM.

In DRAM, data exists in an array (a type of grid containing thousands of neatly arranged bits of data). The array must be electronically refreshed hundreds of times every second in order to maintain its charge and retain its data. If the array loses power, even for a second, all the data it held is erased. That explains why smart computer users take advantage of the Save command at every opportunity.

DRAM technology has undergone many changes since the first PCs hit the market. During the past six years alone, at least half a dozen species of DRAM have had an impact on the memory market. Each new species promises to run faster and more efficiently than its predecessors. Most systems sold today come equipped with SDRAM (synchronous DRAM). See the ďSpecies Of DRAMĒ sidebar for more information about the various species of DRAM.

The shape of DRAM has changed, as well. Originally, DRAM was contained on memory chips that were soldered to the motherboard. These chips held only a small amount of DRAMóan entire system may have contained only 8KB or 16KBóbut consumed a relatively large amount of space. As PCs evolved and the demand for DRAM increased, it became necessary to find a more efficient means of connecting memory chips to the motherboard. The solution came in the form of SIMMs (single in-line memory modules), which were either 30-pin or 72-pin expansion cards that plugged into slots on the motherboard. Later, 72-pin and 168-pin DIMMs (dual in-line memory modules) were developed. DIMMs provide faster data access and more efficient memory storage than SIMMs.

A variant of DRAM has recently become available to consumers. RDRAM, or Rambus DRAM, received its name after the company that developed it. This memory technology is significantly faster than the DRAM installed on most computer systems. RDRAM derives its speed from a proprietary signaling technology that allows the memory modules to operate at speeds of 800MHz (megahertz; one megahertz equals one million cycles per second).

In contrast, typical DRAM runs at speeds between 66MHz and 133MHz. The most notable promoter of RDRAM is Intel. The company produced several chipsets (within a PC, a collection of microchips that augment the CPU by controlling key system components, such as the system clock and caches), including the popular Intel 820 and 840, specifically for the purpose of supporting RDRAM technology.

RDRAM, along with an improved version thatís named DRDRAM (Direct RDRAM), are stored on in-line memory modules called RIMMs. RIMMs are nearly identical to DIMMs, except that RIMMs come in 184-pin models. Because RDRAM runs at such a high speed, the RIMMs also include a built-in heat shield to protect the memory chips from overheating.

A third difference between RIMMs and DIMMs is cost. Youíd pay more than twice as much for a RIMM than for a comparable DIMM. You can find RDRAM on high-end Pentium-based desktop systems produced by Gateway, Dell, Compaq, and other major manufacturers.

Static RAM. SRAM does not need to be continuously refreshed in order to retain its data. This makes it faster and more stable than DRAM. Thereís a price for speed and stability, however, and SRAM costs significantly more than any variety of DRAM. For this reason, SRAM is usually found only in the L1 (level 1) and L2 (level 2) caches, which serve as holding areas between the main memory and the CPU. These caches make up for DRAM deficiencies by holding the most frequently accessed data, thereby expediting processing time.

On newer systems, the L1 cache is a memory chip contained within the CPU. For this reason, L1 cache is often referred to as internal memory. Itís also known as the primary cache because itís the first place the CPU looks for data.

L2 cache, usually located on a separate chip between the CPU and DRAM, is called external memory. Itís also known as the secondary cache because itís the second place the CPU looks for data. L2 cache is larger than L1 cache. A typical PC has between 256KB and 1MB of SRAM.

Peripheral RAM. The systemís main memory and caches arenít the only places youíll find RAM in your system. Video cards and printers often come equipped with their own built-in memory supply. Peripheral RAM temporarily holds data that travels between a peripheral and the rest of the system.

Video RAM. VRAM is the most famous of the peripheral RAMs. Itís usually located on the video card but may be built into the motherboard on some systems. VRAM differs from conventional RAM because two devices can access it simultaneously. This capability lets VRAM receive data from the CPU or the graphics accelerator (the video cardís built-in processor) at the same time it transmits data to the monitor.

Note that the RAM on your video card is not necessarily VRAM. Many video cards use some form of conventional DRAM as video memory. Video cards sold today typically contain between 8MB and 64MB of video memory.

Printer memory. The printer memory, often referred to as the printer buffer, momentarily holds incoming data until the printer is ready to print it. This lets the CPU perform other tasks without waiting for a printer to finish a lengthy print job. The printer buffer usually consists of 4MB or 8MB of DRAM.

  Mode Of Operation. Now that you know a little about the types of memory found in your system, we can explain how they work together with the CPU and the storage devices to process and manage data within the PC. Weíll start with an analogy that sheds a little light on the role of each of these PC components.

Imagine your computer as a giant brain. The primary function of the brain is to think. In a computer, all the thinkingóthe processing and managing of dataóis handled by the CPU. This device lets the brain perform calculations, recognize relationships, and make decisions.

Before it can do any of that, however, the brain must have something to think about. In a PC, that something is data. Data is kept in three areas: ROM, RAM, and the computerís storage devices. Each area plays an important role.

ROM, which provides the most secure and permanent form of data storage in a PC, is like the part of human memory that remembers fundamental information, such as how to walk or speak your native language. This data is built into the infrastructure of the PC and cannot be changed easily.

Two types of RAM provide the computerís short-term memory. SRAM is like the type of short-term memory that lets you remember the first half of a sentence so that you can make sense of the second half. DRAM, on the other hand, is like that type of short-term memory that helps you remember where you parked your car or what you need to buy at the store. In either case, the data is always supplanted quickly by something else.

Storage devices are like long-term memory, holding data for extended periods of time. In a human, this type of memory helps you remember significant information such as your name, the date of your birth, and the name of the pet dog you had as a child. In a computer, the storage devices let you store important documents, financial records, and other data you donít want to lose. You can add information to a storage device or long-term memory at any time.

But these are just analogies. To gain a true understanding of how memory works in your PC, you must explore its complex inner workings. Weíll start with the basics.

All forms of computer memory store data as electronic signals or pulses. Each signal represents a single bit of data. In conventional ROM, the memory chip has built-in, permanent signals. In RAM and newer versions of ROM, the signals are established by an electronic charge. ROM can retain data after the electronic charge ends, but RAM cannot. When RAM loses power, the signals that represent the data are lost. Consequently, the data is lost.

The first time memory comes into play in the system is when you press the power button to turn on the computer. PCs are hardwired to look in the BIOS ROM immediately upon system startup. The BIOS activates various computer components, including the keyboard and disk drives, and initializes the OS. At that point, the OS takes over. It controls how much system DRAM is allocated to each application and to the OS. Integral parts of the OS transfer from the hard drive to DRAM for easy access. Data from any applications contained in the Startup group also transfers to DRAM during this time.

When you open an application, the CPU copies the data from the hard drive and, in the course of processing it, transfers it to DRAM, where it resides while the application remains open. The data is not randomly dumped into DRAM, but is situated in a specific location as determined by the OS. The system bus (the network of wires that carries data between the processor and memory) shuttles data between DRAM, the L1 and L2 caches, and the CPU.

As system capabilities soar, so has the demand for peripheral memory. Video cards such as Creativeís Blaster Annihilator II with 32MB of RAM (an amount that was once unimaginable) are now commonplace.
While all of this is going on, the CPU continuously transmits display data to the video card. If the video card has a graphics accelerator, the accelerator gathers and processes the data. A small portion of video memory is allocated to assist the accelerator in this endeavor. The rest of the video memory is used as a frame buffer, where the display data is organized into individual images before appearing on-screen. The process is similar if you decide to print some data. The CPU transfers the data to the printer buffer, which holds the data and compiles it into a printable page.

When you use the applicationís Save command, the data stored in RAM is copied to the selected storage device. When you close the application, the data should clear out of both DRAM and SRAM. Occasionally, the data isnít cleaned out. If too much old data remains in RAM, the system may lock up unexpectedly.

  History & Development. In 1981, Bill Gates made the bold comment that 640KB of memory ought to be enough for anybody. Today, that much RAM wouldnít be enough to support the Solitaire game that comes built into Windows. Memory has come a long way, not only in terms of quantity, but also in terms of quality and price. There are a few reasons for that.

The first thing that must be understood about the development of memory is that memory must be able to keep up with the processor. This wasnít so much of a problem in the early days, when the processor whirred at the underwhelming rate of 5MHz. But recent technological developments in CPUs have made the task of keeping up a much more difficult thing to do. Consider that it took nearly 25 years for Intel to produce a 150MHz chip, and it took less than two years for them to produce one three times as fast.

The CPU is not an island in the computer. Its capabilities are limited by several things, including the time it takes for data to travel between RAM and the CPU. Your PC may contain a processor capable of running at 950MHz, but if the memory has a maximum capability of 133MHz, your CPU wonít be able to function at its optimum speed. Trying to keep up with the CPU is the reason you see so many different versions of DRAM on the market.

In addition to SDRAM, which is currently the most common form of DRAM, systems sold today may contain DDR SDRAM (double-data-rate SDRAM) and RDRAM for use in current and future systems. Older computers may use FPM DRAM (fast page mode DRAM) and EDO DRAM (extended data output DRAM). See the ďSpecies Of DRAMĒ sidebar for more information about the various species of DRAM.

So, the CPU has fueled the development of memory. But what accounts for the increase in memory required by systems in recent years? You can blame (or credit) the software industry for that. As long as there have been computers, software developers have pushed their products to do more, and to do it faster and more conveniently for end users. You only need to look at the OS progression to see how the demand for more has changed personal computing.

In 1990, most PCs ran the text-based DOS OS. DOS had relatively few built-in functions and made few demands on a system, serving primarily as a platform on which to run computer applications. It was also boring, ugly, and difficult for beginners to use.

Compare that to Windows Me, the OS found on most computers sold today. WinMe is much more than an OS; itís a communications base, an entertainment center, and an office assistant. It looks sharp, supports more than 16 million colors, and plays music and motion video. That type of functionality requires a sophisticated computer system and more memory. Multimedia capabilities in particular have pushed the memory requirements of the PC. All those bells and whistles come with a price.

Fortunately for consumers, the price has decreased precipitously in the last four years, dropping from more than $25 per megabyte in 1995 to a little more than $1 per megabyte in 2001. The reason for the steep price cut is simple: overproduction. Prior to 1995, memory manufacturers could not keep up with demand. The price of RAM went up, the cost of production went down, and memory manufacturers made healthy profits. This caused a reaction in the industry. In 1995, several companies built new fabs (laboratories where memory chips are manufactured) to capitalize on the lucrative memory market. The average fab costs $2.5 billion to build and lasts about three or four years before its technology becomes obsolete. Consequently, a memory manufacturer has to sell a lot of memory fast to recover its costs.

Thatís an easy thing to do when demand and prices are high. But itís not so easy when the market is saturated, as it was after all the fabs began production in 1995. Suddenly, the supply of memory far exceeded the demand, and the price of memory plummeted. The manufacturers, who had invested enormous amounts of capital in the fabs, had no choice but to produce even more memory chips to make up for low prices. This saturated the market even more. As a result, the cost per megabyte of memory dropped more than 90% in the two-year period between 1995 and 1997.

  Futureís So Bright . . . Since 1997, the price of memory has stabilized, and today, the future looks good. Systems will continue to come equipped with ever-greater amounts of DRAM. Prices for memory should remain stable as demand remains high. And new memory technologies, particularly DRDRAM and DDR SDRAM, should eliminate the gap between processor and memory speed.

In the not-too-distant future, system memory will make the big leap to the CPU. Thatís right, all memory eventually will be built into the CPU. This will eliminate concerns about speed because the data-transfer rate (in this case, the time it takes data to travel between the CPU and RAM) will decrease from todayís average of 10ns to 30ns (nanoseconds; a billionth of a second) to almost zero nanoseconds. It may take three years, five years, or 10 years before this becomes a reality. But at some point, those memory modules in your PC will be history, and weíll have to write a new article to tell you how memory works.  

by Jeff Dodd

View the graphics that accompany this article.
(NOTE: These pages are PDF (Portable Document Format) files. You will need Adobe Acrobat Reader to view these pages. Download Adobe Acrobat Reader)

View the chart that accompanies this article.
(NOTE: These pages are PDF (Portable Document Format) files. You will need Adobe Acrobat Reader to view these pages. Download Adobe Acrobat Reader)

View additional graphics that accompany this article.
(NOTE: These pages are PDF (Portable Document Format) files. You will need Adobe Acrobat Reader to view these pages. Download Adobe Acrobat Reader)

Want more information about a topic you found of interest while reading this article? Type a word or phrase that identifies the topic and click "Search" to find relevant articles from within our editorial database.

© Copyright by Sandhills Publishing Company 2001. All rights reserved.