Semiconductor memory

(¦sem·i·kən¦dək·tər ′mem·rē)

(computer science) A device for storing digital information that is fabricated by using integrated circuit technology. Also known as integrated-circuit memory; large-scale integrated memory; memory chip; semiconductor storage; transistor memory.

---

Devices for storing digital information that are fabricated by using integrated circuit technology. Semiconductor memories are widely used to store programs and data in almost every digital system, and have replaced core memory as the main active computer memory.

Many different types of semiconductor memories are used in digital systems to perform various functions—bulk data storage, program storage, temporary storage, and cache (or intermediate) storage. Almost all of the memories are a form of random-access memory (RAM), in which any storage location can be accessed in the same amount of time.

Even though most semiconductor memories can be randomly accessed, they are not all referred to as RAMs. RAMs are memory chips that cannot retain data without power but permit data to be both read from or written into the memory chip's storage locations.

Within the category of read/write RAMs, many subdivisions have been created to satisfy the performance and system architecture requirements of the various applications. Basically there are two types of read/write RAMs—dynamic and static (DRAMs and SRAMs). The terms “dynamic” or “static” refer to the structure of the actual storage circuit (the cell structure) used to hold each data bit within the memory chip. A dynamic memory uses a storage cell based on a transistor and capacitor combination, in which the digital information is represented by a charge stored on each of the capacitors in the memory array. The memory gets the name “dynamic” from the fact that the capacitors are imperfect and will lose their charge unless the charge is repeatedly replenished (refreshed) on a regular basis (every few milliseconds) by externally supplied signals. Static memories, in contrast, do not use a charge-storage technique; instead, they use either four transistors and two resistors to form a passive-load flip-flop, or six transistors to form a flip-flop with dynamic loads for each cell in the array. Once data are loaded into the flip-flop storage elements, the flip-flops will indefinitely remain in that state until the information is intentionally changed or the power to the memory circuit is shut off.

In addition to static and dynamic RAMs, there is an attempt to combine both technologies, thus merging the high storage density of dynamic memory cells with the simplicity of use of static RAMs. Referred to as pseudostatic or pseudodynamic RAMs, these memories include circuits on the chip to automatically provide the refresh signals needed by the dynamic cells in the memory array. Since the signals do not have to be supplied by the external system, the memory appears to function like a static RAM.

There are many other forms of semiconductor memories in use—mask-programmable read-only memories (ROMs), fuse-programmable read-only memories (PROMs), ultraviolet-erasable programmable read-only memories (UV EPROMs), electrically alterable read-only memories (EAROMs), electrically erasable programmable read-only memories (EEPROMs), flash EPROMs, nonvolatile static RAMs (NV RAMs), and ferroelectric memories. Most of these memory types are randomly accessible, but their main distinguishing feature is that once information has been loaded into the storage cells, the information stays there even if the power is shut off.

The ROM is programmed by the memory manufacturer during the actual device fabrication. Here, though, there are two types of ROMs: one is called late-mask or contact-mask programmable, and the other is often referred to as a ground-up design.

As an alternative to the mask-programmable memories, all the other nonvolatile memory types permit the users to program the memories themselves. The fuse PROM is a one-time programmable memory—once the information is programmed in, it cannot be altered.

The birth of the microprocessor in the early 1970s brought with it nonvolatile memory types that offered reusability. Information stored in the memory can be erased—in the case of the UV EPROM, by an ultraviolet light, and in the case of the EAROM, EEPROM, flash EPROM (often referred to as just a flash memory device), nonvolatile (NV) RAM, or ferroelectric memory, by an electrical signal. Then the circuit can be reprogrammed with new information that can be retained indefinitely. All of these memory types are starting to approach the ideal memory element for the computer, an element that combines the flexibility of the RAM with the permanence of the ROM when power is removed. See also Computer storage technology; Integrated circuits; Logic circuits; Microprocessor.

---

A chip that holds programs and data either temporarily (RAM), permanently (ROM, PROM) or permanently until changed (EPROM, EEPROM, flash memory). See memory types and memory module.

A Wafer Full of Memory Chips

Each of the rectangles on this earlier silicon wafer is a four-megabit RAM chip. The wafer is the structural unit that all chips are fabricated on. The chips are cut out and placed into their individual housings. (Image courtesy of Motorola, Inc.)

Memory in the Mid-1960s

They were not chips, but magnetic cores, which this young lady is stringing by hand for the Univac 494 computer. One chip in the wafer in the previous picture above contains more memory than in the entire finished board of 72 squares (core matrices) in this picture. (Image courtesy of Hagley Museum and Library.)

Twelve Magnetic Cores

This shows the details of the wiring of 12 magnetic cores from the Whirlwind computer in the early 1950s. (Image courtesy of The MITRE Corporation Archives.)

Download Computer Desktop Encyclopedia to your iPhone/iTouch

This entry is from Wikipedia, the leading user-contributed encyclopedia. It may not have been reviewed by professional editors (see full disclaimer)