Increasing Functional Specificity over Increasingly Commoditized Hardware - Guide to Cloud Computing for Business and Technology Managers - page 11

chips. These devices came to be called semiconductors . Computer memories
expanded to 2 megabytes of RAM memory, and processing speeds acceler-
ated to 5 million instructions per second (MIPS). The third-generation com-
puters introduced software that could be used by people without extensive
technical training, making it possible for a much larger section of people to
use them in their respective areas in business.
Fourth Generation: Very-Large-Scale Integrated Circuits, 1980-1990
Fourth-generation computers relied on very large-scale integrated circuits
(VLSIC), which were packed with hundreds of thousands, and later millions,
of circuits per chip. These devices came to be called microprocessors . Computer
memory sizes ballooned to over 2 gigabytes (GB) or more, while processing
speeds exceeded 200 MIPS or more. Correspondingly, costs fell precipitously
making possible inexpensive desktop computers that were widely used in
business and everyday life. The fourth generation of computers was charac-
terized by further miniaturization of circuits, increased multiprogramming,
and virtual storage memory.
VLSIC technology has fuelled a growing movement toward
microminiaturization, entailing the proliferation of computers
so small, fast, and cheap that they have become ubiquitous and
almost invisible . For example, many of the intelligent features
that have made automobiles, stereos, toys, watches, cameras, mobiles,
and other equipment easier to use are enabled by microprocessors.
Fifth Generation: Non-von Neumann Architectures, 1990-Present
Fifth-generation computers are based on non-von Neumann architectures
entailing massively parallel processing for handling multimedia data (voice,
graphics, images, and so on). Processing speeds exceed 500 MIPS or more.
Conventional computers are based on the von Neumann architecture, which
processes information serially, one instruction at a time. Massively parallel
computers have a huge network of processor chips interwoven in a com-
plex and flexible manner. As opposed to parallel processing, where a small
number of powerful but expensive specialized chips are linked together,
massively parallel processing (MPP) machines chain hundreds or even thou-
sands of inexpensive, commonly used chips to attack large computing prob-
lems, attaining supercomputer speeds. MPP have cost and speed advantages
over conventional computers because they can take advantage of off-the-
shelf chips to accomplish processing at one-tenth to one-twentieth the cost of
conventional supercomputers.