Architecture (Несколько текстов для зачёта), страница 6

In 1967 Fairchild Semiconductor introduced a single integrated circuit that contained all the arithmetic logic functions for an eight-bit processor. (A bit is the smallest unit of information used in computers. Multiples of a bit are used to describe the largest-size piece of data that a CPU can manipulate at one time.) However, a fully working integrated circuit computer required additional circuits to provide register storage, data flow control, and memory and input/output paths. Intel Corporation accomplished this in 1971 when it introduced the Intel 4004 microprocessor. Although the 4004 could only manage four-bit arithmetic, it was powerful enough to become the core of many useful hand calculators at the time. In 1975 Micro Instrumentation Telemetry Systems introduced the Altair 8800, the first personal computer kit to feature an eight-bit microprocessor. Because microprocessors were so inexpensive and reliable, computing technology rapidly advanced to the point where individuals could afford to buy a small computer. The concept of the personal computer was made possible by the advent of the microprocessor CPU. In 1978 Intel introduced the first of its x86 CPUs, the 8086 16-bit microprocessor. Although 16-bit microprocessors are still common, today’s microprocessors are becoming increasingly sophisticated, with many 32-bit and even 64-bit CPUs available. High-performance processors can run with internal clock rates that exceed 500 MHz, or 500 million clock pulses per second.

The competitive nature of the computer industry and the use of faster, more cost-effective computing continue the drive toward faster CPUs. The minimum transistor size that can be manufactured using current technology is fast approaching the theoretical limit. In the standard technique for microprocessor design, ultraviolet (short wavelength) light is used to expose a light-sensitive covering on the silicon chip. Various methods are then used to etch the base material along the pattern created by the light. These etchings form the paths that electricity follows in the chip. The theoretical limit for transistor size using this type of manufacturing process is approximately equal to the wavelength of the light used to expose the light-sensitive covering. By using light of shorter wavelength, greater detail can be achieved and smaller transistors can be manufactured, resulting in faster, more powerful CPUs. Printing integrated circuits with X-rays, which have a much shorter wavelength than ultraviolet light, may provide further reductions in transistor size that will translate to improvements in CPU speed.

Many other avenues of research are being pursued in an attempt to make faster CPUs. New base materials for integrated circuits, such as composite layers of gallium arsenide and gallium aluminum arsenide, may contribute to faster chips. Alternatives to the standard transistor-based model of the CPU are also being considered. Experimental ideas in computing may radically change the design of computers and the concept of the CPU in the future. These ideas include quantum computing, in which single atoms hold bits of information; molecular computing, where certain types of problems may be solved using recombinant DNA techniques; and neural networks, which are computer systems with the ability to learn.

Computer Memory

Computer Memory, device that stores data for use by a computer. Most memory devices represent data with the binary number system. In the binary number system, numbers are represented by sequences of the digits 0 and 1. In a computer, these numbers correspond to the on and off states of the computer’s electronic circuitry. Each binary digit is called a bit, which is the basic unit of memory in a computer. A group of eight bits is called a byte, and can represent decimal numbers ranging from 0 to 255. When these numbers are each assigned to a letter, digit, or symbol, in what is known as a character code, a byte can also represent a single character.

Memory capacity is usually quantified in terms of kilobytes, megabytes, and gigabytes. The prefixes kilo-, mega-, and giga-, are taken from the metric system and mean 1 thousand, 1 million, and 1 billion, respectively. Thus, a kilobyte is approximately 1000 (1 thousand) bytes, a megabyte is approximately 1,000,000 (1 million) bytes, and a gigabyte is approximately 1,000,000,000 (1 billion) bytes. The actual numerical values of these units are slightly different because they are derived from the binary number system. The precise number of bytes in a kilobyte is 2 raised to the 10th power, or 1,024. The precise number of bytes in a megabyte is 2 raised to the 20th power, and the precise number of bits in a gigabyte is 2 raised to the 30th power.

Computer memory may be divided into internal memory and external memory. Internal memory is memory that can be accessed directly by the central processing unit (CPU)—the main electronic circuitry within a computer that processes information. Internal memory is contained on computer chips and uses electronic circuits to store information. External memory is memory that is accessed by the CPU via slower and more complex input and output operations. External memory uses some form of inexpensive mass-storage media such as magnetic or optical media. See also Information Storage and Retrieval.

Memory can further be distinguished as being random access memory (RAM), read-only memory (ROM), or sequential memory. Information stored in RAM can be accessed in any order, and may be erased or written over depending on the specific media involved. Information stored in ROM may also be random-access, in that it may be accessed in any order, but the information recorded on ROM is permanent and cannot be erased or written over. Sequential memory is a type of memory that must be accessed in a linear order, not randomly.

Random access memory is the main memory used by the CPU as it processes information. The circuits used to construct this main internal RAM can be classified as either dynamic RAM (DRAM), or static RAM (SRAM). In DRAM, the circuit for a bit consists of one transistor, which acts as a switch, and one capacitor, a device that can store charge. The bit 1 is stored in DRAM by a charged capacitor, while the bit 0 is stored in DRAM as an uncharged capacitor. To store the binary number 1 in a DRAM bit location, the transistor at that location is turned on, meaning that the switch is closed, which allows current to flow into a capacitor and charge it up. The transistor is then turned off, meaning that the switch is opened, which keeps the capacitor charged. To store a 0, charge is drained from the capacitor while the transistor is on, and then the transistor is turned off. To read a value in a DRAM bit location, a detector circuit determines whether charge is present or absent on the relevant capacitor. Because capacitors are imperfect, charge slowly leaks out of them, which results in loss of the stored data. Thus, the computer must periodically read the data out of DRAM and rewrite it by putting more charge on the capacitors, a process known as refreshing memory.

In SRAM, the circuit for a bit consists of multiple transistors that continuously refresh the stored values. The computer can access data in SRAM more quickly than DRAM, but the circuitry in SRAM draws more power. The circuitry for a SRAM bit is also larger, so a SRAM chip holds fewer bits than a DRAM chip of the same size. For this reason, SRAM is used when access speed is more important than large memory capacity or low power consumption.

The time it takes the CPU to read or write a bit to memory is particularly important to computer performance. This time is called access time. Current DRAM access times are between 60 and 80 nanoseconds (billionths of a second). SRAM access times are typically four times faster than DRAM.

The internal memory of a computer is divided into locations, each of which has a unique numerical address associated with it. In some computers an address refers directly to a single byte in memory, while in others, an address specifies a group of four bytes called a word. Computers also exist in which a word consists of two or eight bytes.

When a computer executes a read instruction, part of the instruction specifies which memory address to access. The address is sent from the CPU to the main memory (RAM) over a set of wires called an address bus. Control circuits use the address to select the bits at the specified location in RAM. Their contents are sent back to the CPU over another set of wires called a data bus. Inside the CPU the data passes through circuits called the data path. In some CPUs the data path may directly perform arithmetic operations on data from memory, while in others the data must first go to high-speed memory devices within the CPU called registers.

Read-only memory is the other type of internal memory. ROM memory is used to store the basic set of instructions, called the basic input-output system (BIOS), that the computer needs to run when it is first turned on. This information is permanently stored on computer chips in the form of hardwired electronic circuits.

External memory can generally be classified as either magnetic or optical, or a combination called magneto-optical. A magnetic storage device uses materials and mechanisms similar to those used for audio tape, while optical storage materials use lasers to store and retrieve information from a plastic disk. Magneto-optical memory devices use a combination of optical storage and retrieval technology coupled with a magnetic media.

Magnetic tape is one form of external computer memory, but instead of recording a continuous signal as with analog audio tape, distinct spots are either magnetized or demagnetized on the tape, corresponding to binary 1s and 0s. Computer systems using magnetic tape storage devices employ machinery similar to that used with analog tape: open-reel tapes, cassette tapes, and helical-scan tapes (similar to video tape).

Another form of magnetic memory uses a spinning disk coated with magnetic material. As the disk spins, a sensitive electromagnetic sensor, called a read-write head, scans across the surface of the disk, reading and writing magnetic spots in concentric circles called tracks.

Magnetic disks are classified as either hard or floppy, depending on the flexibility of the material from which they are made. A floppy disk is made of flexible plastic with small pieces of a magnetic material imbedded in its surface. The read-write head touches the surface of the disk as it scans the floppy. A hard disk is made of a rigid metal, with the read-write head flying just above its surface on a cushion of air to prevent wear.

Optical external memory uses a laser to scan a spinning reflective disk in which the presence or absence of nonreflective pits in the disk indicates 1s or 0s. This is the same technology employed in the audio compact disc (CD). Because its contents are permanently stored on it when it is manufactured, it is known as compact disk-read only memory (CD-ROM). A variation on the CD, called compact disk-recordable (CD-R) uses a dye that turns dark when a stronger laser beam strikes it, and can thus have information written permanently on it by a computer.

Magneto-optical (MO) devices write data to a disk with the help of a laser beam and a magnetic write-head. To write data to the disk, the laser focuses on a spot on the surface of the disk heating it up slightly. This allows the magnetic write-head to change the physical orientation of small grains of magnetic material (actually tiny crystals) on the surface of the disk. These tiny crystals reflect light differently depending on their orientation. By aligning the crystals in one direction a 0 can be stored, while aligning the crystals in the opposite direction stores a 1. Another, separate, low-power laser is used to read data from the disk in a way similar to a standard CD-ROM. The advantage of MO disks over CD-ROMs is that they can be read and written to. They are, however, considerably more expensive than CD-ROMs.

Since the inception of computer memory, the capacity of both internal and external memory devices has grown steadily at a rate that leads to a quadrupling in size every three years. Computer industry analysts expect this rapid rate of growth to continue unimpeded. Computer scientists consider multigigabyte memory chips and terabyte-sized disks real possibilities. Research is also leading to new optical storage technologies with 10 to 100 times the capacity of CD-ROMs produced in the 1990s.