Information Storage Technologies: Types of Computer Memory, RAM vs ROM, and External Storage

Computer memory, or storage devices, fall into three groups: internal RAM (random access memory), permanent ROM (read-only memory), and external storage such as hard disks, optical discs, and flash media. Each type serves a distinct role — RAM holds the data the processor works with right now, ROM holds the firmware that starts the machine, and external storage keeps your files when the power is off.

• RAM — random access memory, the fast working memory the processor reads and writes during operation;
• ROM — permanent (read-only) memory that retains start-up data and programs without power;
• External storage — hard disk drives, floppy disks, compact discs, magnetic tape drives, magneto-optical discs, and flash cards.

Data and programs are usually stored on a hard disk. Programs launched for execution, and documents accessed by a running program, are loaded into RAM because the processor's access time to RAM is far shorter than its access time to an external storage device. The permanent memory device stores the service data and programs needed to start the computer the moment the power is turned on.

What is RAM and what does it do?

RAM (Random Access Memory) holds the information the processor is working with at a given moment. It is the computer's short-term working memory: when you open a program or document, the system copies it from the hard disk into RAM so the processor can reach it quickly.

RAM consists of microchips mounted on a small board that slots into a dedicated connector on the computer's motherboard. Data in RAM persists only while the computer is connected to power and is lost the instant power is disconnected — this is why unsaved work disappears after a sudden shutdown.

How is RAM structured internally?

The internal structure of RAM can be pictured as a rectangular matrix of cells, each able to store one bit — a single 1 or 0. Data moves between the processor and RAM along a dedicated bus, sometimes called the information highway, and one such exchange is known as a bus cycle.

The number of data bits the processor transmits or receives in a single bus cycle is called the bus width. RAM is produced as expansion cards, or modules, that insert into designated sockets on the motherboard.

RAM modules have appeared in several form factors over the years:

• SIMM (Single In-line Memory Module) — boards with a single-row arrangement of memory chips; obsolete and almost never used today.
• DIMM (Dual In-line Memory Module) — originally boards with one and two rows of memory chips; the basis of modern desktop memory.
• RIMM (Rambus In-line Memory Module) — the module used by Rambus DRAM systems.

What are the types of RAM chips?

RAM chips come in two broad families — dynamic (DRAM) and static (SRAM) — that differ in speed, cost, and how they hold a bit. RAM is used across a wide range of personal computer devices, from video cards to laser printers, and in those general roles the chips are almost always dynamic RAM (DRAM).

Static RAM (SRAM) is the faster, more expensive type used for buffer or cache memory in hard disks and CD-ROM drives, and for the level-one and level-two cache built onto processors. Its capacity is relatively small — hundreds of kilobytes against tens or hundreds of megabytes of DRAM — but it operates roughly ten times faster.

A static memory cell (SRAM) is an electronic device called a trigger (flip-flop) that rests in one of two stable states: logical zero or logical one. The trigger holds its state until the power is removed or a signal arrives instructing it to switch.

A dynamic memory cell (DRAM) holds its state in the charge of a tiny capacitor: while the charge is present the cell reads as a one, and with no charge it reads as a logic zero. Because the charge leaks away within a few milliseconds, DRAM must be refreshed periodically by reading and rewriting the cells. The cells of dynamic memory are arranged in a matrix called a bank.

RAM ships as chips assembled into memory modules. A few years ago, 72-pin SIMM modules were common and had to be installed in pairs, since each module represented half of a standard memory "bank". In 1998, 168-pin DIMM modules reached the market and could be installed one at a time. A single DIMM module can hold from 1 MB to 512 MB of RAM, though in practice the common sizes were 128 MB and 256 MB.

For computers running Windows 98, 2000, Me, or XP, 128 MB of RAM was the practical minimum and 256 MB or more the optimal amount. Most motherboards of the era offered three slots for memory modules. Modules of different sizes could be mixed — for example, two 64 MB modules and one 128 MB module — but it was best to match access speed (for instance 7 ns) and use modules from the same manufacturer.

What are the main types of RAM (SDRAM, DDR, RDRAM)?

The main RAM types are SDRAM, DDR SDRAM, and Direct Rambus DRAM, each a step up in bandwidth. They share the dynamic-memory principle but differ in how they synchronise with the system and how much data they move per clock cycle.

SDRAM (Synchronous DRAM) is synchronised with the system timer that the central processor controls. It has a data access time of 6 to 9 ns and a bandwidth of 256 to 1000 MB/s, and its key specification is the maximum operating bus frequency. Three modifications were sold:

• PC66 — 9-ns modules running on a system bus up to 83 MHz (base frequency 66 MHz); now obsolete.
• PC100 — 8-ns modules running up to 110–120 MHz (base frequency 100 MHz).
• PC133 — 7-ns modules supporting bus frequencies up to 150 MHz (base frequency 133 MHz).

DDR SDRAM (Double Data Rate DRAM) is the next generation of SDRAM. It works on the same principles but transfers data on both edges of the synchronised clock signal, which doubles the effective transfer rate. Compared with conventional SDRAM, DDR SDRAM reaches a bandwidth of about 3.2 GB/s at an access time of 5–6 ns. Its effective bus frequency runs around 600–700 MHz, so even at standard clocks of 100 and 133 MHz its performance is roughly twice as high. DDR modules are labelled by bandwidth rather than bus frequency: PC1600, PC2100, PC2700, and PC3200 correspond to working bus frequencies of 100, 133, 166, and 200 MHz respectively.

DRDRAM (Direct Rambus DRAM, or RDRAM) was developed by Rambus Inc. It uses a multifunctional data-exchange protocol that moves data over a simplified bus running at very high frequency, making it a system-level integrated technology. RDRAM supports bus frequencies of 800 MHz for PC800 modules and up to 1066 MHz for PC4200 modules, a memory access time of 4 ns, and transfer rates up to 6 GB/s. Intel's earliest chipsets, the 840 and 820, were built around it. RDRAM modules cost several times more than SDRAM and DDR SDRAM, but motherboards for Intel Pentium 4 processors (i850, i850E, and similar) were designed to use it.

What is cache memory and how do L1 and L2 work?

Cache memory is a small, very fast buffer that sits between the processor and RAM to smooth the speed gap between them. Because the processor runs faster than RAM can respond, cache holds the most frequently used data close to the core so the processor rarely has to wait on slower main memory.

CPU cache is organised in two levels: first-level cache (L1) and second-level cache (L2). In early processors — Pentium, Pentium II, and Pentium III (Katmai) — the L2 cache was sometimes mounted on the motherboard as a separate chip.

The processor looks for data first in the _L1 cache, then in the _L2 cache, and only afterwards in RAM. In later processors — Pentium III, Celeron (Coppermine, Tualatin), Pentium IV, Athlon (Thunderbird), Athlon XP (Palomino, Thoroughbred), and Duron — the L2 cache was built directly into the processor. Integrating the cache this way gives the processor far faster access to second-level cache than it has to RAM.

What is virtual memory (the swap file)?

Virtual memory is an area on the hard disk used as an extension of RAM when physical memory runs short. When a program needs more memory than is physically installed, the operating system treats part of the disk as additional RAM so the program can still run. Under MS-DOS such oversized programs simply would not start; operating systems like Windows solved the problem with virtual memory.

Virtual memory — also called the swap file or paging file — has a longer access time than real RAM, but its presence lets you run programs that would otherwise be impossible. By default the C: drive hosts the swap file, and its maximum size equals the free space on that drive. For this reason you should always keep roughly 200–500 MB free on the system drive.

What is permanent memory (ROM)?

Permanent memory, or ROM (Read Only Memory), holds data that cannot be erased or overwritten under normal use and is retained indefinitely even with the power off. ROM stores the firmware the computer needs to boot, which is why it must survive shutdowns.

How is ROM designed and programmed?

ROM is realised as chips installed on the computer's motherboard, with information written once at the manufacturing stage. By the method used to record data, ROM chips fall into two categories:

1. Mask ROM — written once during the production of the chip itself.
2. Programmable ROM, which subdivides further into:
   • One-time programmable ROM (PROM — Programmable ROM);
   • Reprogrammable ROM (EPROM — Erasable PROM), including the electrically erasable variant (EEPROM — Electrically Erasable PROM).

The contents of an EEPROM can be changed by applying a special electrical charge to the chip's pins, which is what makes firmware updates possible without replacing hardware.

What is FLASH memory and how did it develop?

FLASH memory is a non-volatile storage type that keeps data without power yet can be rewritten like RAM, combining the best traits of ROM and RAM. Intel Corporation invented it in 1988, and the name comes from its original erase method — wiping the whole chip at once, in a "flash". Modern chips have shed that limitation and erase byte-by-byte or page-by-page instead.

FLASH memory grew out of EEPROM, driven less by technical refinement than by the goal of cutting cost. The first flash-based chip, with a capacity of 256 KB, was used in Hewlett-Packard medical equipment. The technology gained wide popularity in the 1990s: Toshiba announced the Solid-State Floppy Disk Card (SSFDC), SanDisk introduced CompactFlash in 1994, and Sony released the Memory Stick in 1998.

FLASH chips of that era held 32 MB to 1024 MB, came in several supply voltages, and could withstand more than a million write/read cycles, with manufacturers guaranteeing data retention for at least ten years. Like RAM, FLASH memory lets you overwrite stored data; unlike RAM, it keeps that data through a power failure. FLASH modules are built on EEPROM technology, and their read and write times match those of dynamic RAM. A further advantage is that FLASH memory draws power only at the moment of reading or writing.

Several flash card formats were popular around 2000, each tuned to particular devices:

1. SmartMedia Card (SMC) — announced in 1998 at 45×37 mm, compatible with cameras and devices from Epson, Agfa, Minolta, Olympus, Ricoh, Sanyo, Fuji Photo Film, and Toshiba.
2. MultiMediaCard (MMC) — the smallest format at 24×32 mm.
3. Secure Digital Card (SD) — a further development of the MMC, used in the Palm and Toshiba Pocket PC families.
4. CompactFlash (CF) — used in digital cameras, MP3 players, and PDAs.
5. Memory Stick (MS) — first used only in Sony digital cameras, MP3 players, and PDAs, later adopted more widely.

To explore related hardware and software topics, see our explainer on what software is in the PC section.