Computer Hardware

1.1 History

The notion of a mechanical calculator is due to Charles Babbage. In 1821 he proposed his Difference Engine consisting of a set of linked adding mechanisms. Its primary purpose was to manipulate polynomials (by computing finite differences). He built a prototype which could compute second order differences. This machine could compute values of quadratic polynomials. Although he designed one capable of dealing with a sixth degree polynomial, he never built one. His Analytical Engine is the forerunner of the architecture of modern computers. It had a storage unit, arithmetic "mill" or unit and punched card input/output. The building technology was based on gears and linkages. According to his estimation, an addition could be done in a second, a multiplication in a minute. Babbage died in 1871. His designs were never implemented since his ideas were far too advanced for his time.

First Generation Machines (1945-58)

These machines used thermionic valves. Two prototypes were built at the University of Pennsylvania (EDVAC and ENIAC). These were huge machines, 80 feet long, 8 feet high with 17,000 vacuum tubes. A major characteristic of a value is known as the gate delay time(GDT). The GDT is the amount of time taken by a signal to travel from the input of one logic gate to the input of the next. The typical GDT for such a machine was a microsecond (10^-6 seconds). The major disadvantages were that they required extraordinary amount of room, dissipation of large amount of heat, and frequent break down! A typical commercial machine is the IBM 704.

Second Generation Machines (1959-65)

In the second generation machines, the vacuum tube gave way to the transistor. The transistor was developed by American Physicists Bardeen, Brattain and Shockley of AT&T Bell Labs. They shared the Nobel prize in 1956 for their invention. A transistor is much smaller than a vacuum tube and so can house more active components. They were very reliable, produced very little heat. Computers based on the transistor were very compact. The GDT for such machines had been reduced to 0.3 microseconds. A typical machine of this generation is IBM 7090. The minicomputer (DEC) also belongs to this generation.

Third Generation Machines (1965-1975)

A new component technology was born in 1965 - solid state integrated circuits (invented by Jack Kelby at Texas Instruments). An IC is the size of an ice cube that contains hundreds of miniature transistors). The chief characteristic of these circuits is the silicon chip where several components were combined in a single wafer. In the beginning there were only few gates were placed on such a chip with a GDT of 10 nanoseconds (10^-9 seconds). By 1975 this was improved to 1 nanosecond by increasing the number of gates. Typical machines of this generation are the hugely successful IBM 360 series, Burroughs 6500 and the UNIVAC 1108.

Fourth Generation Computers (1980 - )

This age belongs to the microchip, or more commonly known as VLSI - very large scale integrated circuit. A microchip is the size of a postage stamp containing hundreds of thousands of electronic components. This is the age of the micro or "personal" computer which is still evolving.

The GDT had thus improved from 1 microsecond to 1 nanosecond (1000 fold decrease) in 30 years. Typical machine performances from these generations are:

1st generation: 100 arithmetic operations per second (EDSAC)

2nd generation: 100,000 operations per second (IBM 7090)

3rd generation: 10 million operations per second (10 mflops, IBM 360)

1.The von Neumann Architecture

All the computers designed in these generations, including the Babbage Analytical Engine, were based on essentially one model of architecture: the von Neumann model. This consisted essentially of one each of the following:

I/O : input/output device

MEMORY location for storage of data and instructions

CU : control unit for instruction interpretation

ALU : arithmetic and logical unit for processing data.

(Dotted line : Data, Solid Line: Control)

In this model, under the control of the control unit, the following operations are repeated over and over until it reaches a "stop" instruction:

1. Read an instruction from memory;

2. Read any data required by instructions from memory;

3. Perform the operation(s) on the data;

4. Store results in memory;

5. Go to 1.

The hardware of modern day computers acts in much more complex manner. However, for purposes of understanding, this simple model suffices. The important fact to note is that processing speed of the machine is limited by the rate at which instructions and data can be transferred from memory to processing unit. This narrow connection between instructions and data held in memory and the single processing unit forms the so called von Neumann's bottleneck.

In spite of the bottleneck, each succeeding generation of computer evolution, the processing speed of the machines, measured by the number of operations per second had improved by an order of 10⁵. There has been a real need for faster and faster machines to solve larger and more demanding problems.

MEMORY AND THE BUS SYSTEM

Operation of a computer requires a means of storing binary encoded information. This includes representations of numerical or other data as well as machine language programs as discussed above. Computer memory serves this function.

Binary data within the computer is packaged into a units of a fixed number of bits. Almost always, this size is 8 bits so we will assume that the computer stores binary data as a sequence of bytes. In order to distinguish one byte from another, each memory location is associated with a specific binary string, its "address".

Example Suppose a computer uses 4 bit addresses, what are the maximum number of bytes that can be stored in memory at any one time?

Solution: Since there are 2⁴ different 4 bit binary codes, 16 bytes of memory can be accessible. Figure shows a diagram of memory in which every byte is associated with a 4 bit address. The address to the first byte is 0000b, the address to the second byte is 0001b, and so on through the maximum number of storage locations available. The address to the last storage location is 1111b.

Example (a) What is the largest possible memory size for a computer which uses 10 bit addresses? (b) What about 20 bit addresses?

Solution The solution to (a) is 2¹⁰ = 1024 bytes (1 KiloByte or 1 KB) and the solution to (b) is 2²⁰ = 1032576 bytes (1 MegaByte or 1 MB)

The bus system is composed of three components: the address bus, the data bus and the control bus. Data travels between the CPU, memory and I/O over the data bus. The CPU uses the address bus to specify a unique address corresponding to a specific memory location or a specific input/output device. The CPU uses the control bus to indicate whether the data is to be go into or out of the CPU. The CPU also uses the control bus to synchronize the actual data transfer between the CPU, memory and I/O.

When the CPU performs an operation, as directed by a program instruction, memory may serve as either the source or destination operand. Accessing memory by the CPU is accomplished by the following steps: (i) The CPU uses the address bus to communicate the address of the relevant memory location to computer memory. (ii) The CPU uses the control bus to send a control signal to computer memory specifying whether access is to be a "READ" or a "WRITE". By convention, movement of data to the CPU is always referred to as a READ; movement of data from the CPU is always referred to as a WRITE. (iii) Data is placed on the data bus either by memory (a READ) or by CPU (a WRITE) (iv) Data is retrieved from the data bus by the other component.

There are two kinds of computer memory. Random Access Memory (RAM) is a general sort of memory as discussed above. Read Only Memory(ROM) differs from RAM in that a binary data in this sort of memory is placed there by the "factory" and cannot be changed by the CPU. RAM is sometimes called volatile memory since when the computer is turned off this memory disappears. When the computer is turned again , any data in RAM is lost. ROM is sometimes called non volatile memory since it always contains the same data. When a computer is first powered up, the program stored at a particular address is always executed first. This must be the address of a location in ROM otherwise the computer will behave unpredictably.

REMARKS:

(1) Memory can be accessed as a byte by specifying its byte address. However, some computers allow accessing multiple consecutive bytes by specifying a byte address. Either the lower byte address or the higher byte address can be specified depending on the computer.

(2) Memory may sometimes be organized into "segments". Segmentation gives a logical separation of machine instructions and data.

(3) All computers use an area of RAM as a "stack". Access to the stack is normally restricted to a "push" operation which writes data into the stack and a "pop" operation which retrieves from the stack the last value pushed. This type of access is referred to as "last-in-first-out". These stack operations may be invoked explicitly by the programmer and may be used by the CPU in some situations.

(4) Most modern day computers organize memory in a hierarchical fashion as registers (the lowest level of the hierarchy), cache memory, main memroy(RAM/ROM), and secondary memory(the highest level of the hierarchy). At lower levels in the hierarchy, memory is more expensive but is much faster. Economic factors restrict the amount of memory at the lower levels of the hierarchy. Depending on the particular type of computer used, the distinctions between various levels of the hierarchy may or may not be visible to the programmer.