|BETTER, QUICKER, FASTER
IN PART ONE AND TWO OF THIS SERIES PETER HAYES HAS
EXPLORED THE KEY FACTORS THAT GOVERN COMPUTER PROCESSING
TODAY - IN PART THREE - HE ROUNDS
UP THE SERIES BY LOOKING BACK AT THE MILESTONES OF CHIP
DESIGN AND THEN FORWARD TO THE FUTURE OF THE MEDIUM.
In part one and two of this series
I have tried to emphasis that computing is very much a
team effort, and that while faster processors play an
important role they are not the whole story.
I also outlined the fact that the computer chip industry
is becoming more and more competitive with leading player
Intel now feeling the pressure from companies such as
AMD, Cyrix and IBM.
Today, however, we examine the steps that have been taken
to reach this point in history and have an educated guess
at the close and distant future.
In computer design circles the 1970 "Intel
4004" will always have a special place. Today the
"4004" is acknowledged as being the first
example of a "general purpose" microprocessor.
Before that integrated chips had been designed to serve
only one specific purpose.
Designed by Ted Hoff and Frederico Faggin the 4004 chip
was used in the very first generation of (very expensive)
silicon-based calculators. It contained the equivalent of
2300 single transistors and was the tiny acorn from which
chips such as the Pentium Pro range would grow.
Today modern chips can contained over 5.5 million
transistors - and given that newer models have caches and
co-processors built-in - the number seems certain to
grow. However, while these figures give a good
rule-of-thumb as to how powerful the processors are,
further improvements can still be achieved by further
Progress was quick in those early days and by 1974 Intel
had come up with the 8080, the first general purpose chip
designed for "full computer" use and already
twenty times faster than the 4004 family. This 8-bit chip
found its way in to many kit computers - including the
famous Altair - and was more notable for bringing the
home computer a step closer.
In 1979 Intel produced the 8088 - perhaps the biggest
breakthrough chip of all time. This 16-bit processor
drove the first IBM PC which was soon cloned by a wide
variety of producers. It's introduction sent Intel share
prices in to orbit.
The next milestone was the 80286 - the first of the
"286" processors - in 1982. The chip contained
the equivalent of 130,000 transistors and was powered by
a 12 MHZ clock. Natural progress followed with the
introduction of the 80386 chip (the first of the 386
models) in 1985 and 80486 (the first of the 486 models)
in 1989. The 386 model featured 275,000 transistors, but
this number grew to more than a million for the 486
The next breakthrough came in 1993 with the first Pentium
range. While containing more than 3 million transistors
the chip was better designed to take into account
supporting graphics and communications applications. In
short Intel had looked more closely at the role of
computers in the modern world rather than simply go for
The Pentium Pro came out in 1995 and featured what Intel
called "dynamic instruction execution" which
means that the maths was streamlined before being
performed. The chip also featured the first in-built
cache. The Pentium Pro now features more than 5.5 million
While it may be crude to say that other producers have
been playing catch-up, this is essentially what has been
happening within the chip industry. Intel have even tried
to claim, through the courts, that their rivals are
merely "copying" there products - although
While many people view the games industry as a
bit-of-a-joke many great strides forward have come about
by the games industry - who have been at the forefront of
producing high standards in graphic images. Commercial
areas such as virtual reality and flight simulation have
borrowed heavily from the games industry.
While I've outlined most of the key problems of computer
speed in parts one and two of this series, the key debate
has been about affordable and practical computing.
For those with bottomless pockets many computer problems
can be overcome: Computers from Cray are - at their very
heart - simply endless rows of processors wired together
like a team of horses. In technical circles these are
called massively parallel systems or MPPs.
The problem with this type of computer is not only the
expensive component parts, but having to invest far more
in software that has to be more structured and
Nevertheless, as outlined in parts one and two, the
next-step-forward is to provide better support for the
main CPU through co-processors. In short, the same sort
of idea, but on a much smaller and more automatic scale.
To continue our central idea of computing being separate
parts held together by a central theme, we must consider
the central demands of computing.
Computer hardware manufactures can only provide
components that their customers want. A chip that can
perform a record amount of floating-point maths (per
second) will be useless unless there is a reasonable
commercial market for it. In other words the designers
have to design for the commercial market not for the
Equally important is that computers can be improved
simply by being more focused and targeted to the purpose
that it has to perform. If the computer is a games
console, it is obvious that the user wants fast screen
updates and multi-channel sound - things that the
"serious" computer user might not.
In certain cases there are components - such as hardware
caches - that might actually hinder efficiency when
running certain pieces of software, because the design of
the software doesn't gain any advantages from having
them. Or else the cache is too small or too large for the
individual application. Therefore I'd be happier
describing these functions as "most of a good
thing" rather than "all of a good thing."
In more simple terms the design of a computer can never
be perfect. There will always be debates as to whether a
particular added function helps or hinders in the world
it is likely to encounter.
The biggest stone wall facing chip manufacture is the
nature of electricity itself. There is a built in limit
to how quickly the central electrons can travel so chips
cannot simply become faster and faster without end. Some
experts say that the future of computing lies in the use
of semiconductor lasers and memory based in chemicals -
but this will require a breakthrough that will swamp all
those that have gone before.