Moore's Law and the Integrated Circuit - engweb.info

amountdollΗλεκτρονική - Συσκευές

2 Νοε 2013 (πριν από 3 χρόνια και 9 μήνες)

169 εμφανίσεις

Uni versity of Greenwich


EADS, DSE

Mark Clements

Page
1

11/2/2013

Moore’s Law and the Integrated Circuit

Introduction

Integrated circuits (ICs) are found at the heart of all modern electronic equipment
from the mobile telephone to the personal computer. These devices provide the
circuitry required to operate our digital
devices. At the heart of these ICs are
miniature components connected by even smaller electrical paths. The interface to
these devices is by the many ‘legs’ on the exterior of the IC itself. Ever since the
introduction of the first integrated circuit, the
complexity of successive integrated
circuit has constantly increased.

Development of the Integrated Circuit

In early 1961, work was under way on creating the technology to make integrated
circuits on silicon wafers to replace the discrete components that h
ad been used to
create all electronic circuits up to that point.



Capacitors


Resistors



Transistors


The problem with discrete components was that the number of components that are
required to produce complex circuits produces the chance of errors i
n connection
(poor solder joints) and component value errors, placing errors and also the chance of
including dead components. Additionally, the cost of producing such circuits was
high owing to the individual component cost and also the cost of soldering
them
together onto the circuit board.


When components are placed onto the circuit board, the ‘legs’ need to be bent to the
correct shape and the excess ‘leg’ needs to be cut off. The differing length signal
paths that were produced using these separate co
mponents tended to limit the rate at
which these circuits could be run.


It was realised during the 1960s that a whole circuit consisting of many transistors,
resistors and capacitors could be built on a wafer of silicon using masking and
‘spraying’ techn
iques. In this way, the entire circuit could be produced
en masse

and
each of the circuits should be identical to the next. The size and weight would reduce
and the signal paths between the components would be drastically reduced. This
would improve the da
ta rate that a circuit could be run at and also improve the
matching (impedance and data rate) of each circuit with any other in the whole
system.


Uni versity of Greenwich


EADS, DSE

Mark Clements

Page
2

11/2/2013

A separate effect of manufacturing the components on one wafer of silicon at the
same time was to decrease
the cost of the finished circuit. This has led to cheap every
-
day items such as mobile telephones and hand
-
held computers that are easily
affordable.


Another major factor that accelerated the research into integrated circuits was the size
and weight that
the whole circuit would have if it were built entirely from discrete
components. With work on getting man into space, lightweight and small electronic
components would be needed to help with the extremely complex calculations that
would be needed during su
ch a flight.


The power consumption of a circuit made with discrete devices is very much higher
than the consumption of an IC. IC technology has the ability to make mobile devices
that will run for very long periods without requiring a new set of batteries
.

Moore’s Law

By 1965, Gordon Moore, the
director of the Research and Development Laboratories
at Fairchild Semiconductors
observed an exponential growth in the number of
transistors per integrated circuit. He noticed that the number of devices on a silico
n
chip increased regularly and predicted that this trend would continue for at least the
next 10 years. His observations were based on the current and previous work that had
been done on placing multiple devices on a single substrate.


Although Gordon Moo
re never said that the number of transistors would double
approximately every 18 months, this is how the law has been accepted.


By extrapolating his observations, he was able to make assumptions about devices
that were yet to be built. His predictions inc
luded personal computers and high data
rate switched telecommunications systems


all made possible by the cramming of
more and more devices on a single wafer of silicon.


Uni versity of Greenwich


EADS, DSE

Mark Clements

Page
3

11/2/2013

Below can be seen a graph of the trend that Moore predicted (courtesy of
http://www.intel.com/research/silicon/mooreslaw.htm
)


The data that was used to draw this graph can be found in the
appendix

to this
document.


Integrated Circuit Te
chnological Advances

At the outset of integrated circuit production, a technique known as Small Scale
Integration (SSI) was able to combine around 10 discrete components (transistors
diodes & resistors) onto around 5mm square of silicon substrate.


Develop
ment of SSI led to Medium Scale Integration (MSI), then Large Scale
Integration (LSI) with many thousands of components in the same area of silicon.


Very Large Scale Integration (VLSI) provided the means to implement around 1
million components per chip.

VLSI led to 4th generation computers and the microcomputer.

Current technology can produce silicon wafers with over 100 million components per
chip. Intel's Dothan has around 140 million components on the wafer (2007).

Why does Moore’s Law exist?

By look
ing at the graph presented earlier, it is easy to see that the trend predicted by
Moore has remained fairly constant. An interesting question to ask is why has the law
remained fairly constant for almost forty years?


Once Moore had made his observation an
d then made it public, did this force to
manufacturers of IC devices to try to keep up with his prediction? It would be
interesting to theorise whether Moore’s Law would have manifested itself if he did
not make public his famous observation.

Uni versity of Greenwich


EADS, DSE

Mark Clements

Page
4

11/2/2013


Perhaps the

competition between IC manufacturers would have led to the trend
anyway. Currently, the three main players in the IC market are Intel, Cyrix and AMD.
These three companies rival each other to produce the ‘best’ device that they can and
are constantly stri
ving to produce a cheaper, smaller, faster, more feature
-
rich IC than
their opponents. Is this the driving force behind the Law?


Another way to examine the problem here is to consider that the advances required in
the IC industry require much input from c
omputer derived simulation and design.
Does each successive advance in IC technology give the manufacturers ever more
powerful tools to carry out their designs for the next IC design?


Consumers may be responsible for the trend. Is it because there is a de
mand for
smaller, faster, lighter, lower consumption devices that manufacturers continue to
produce ever better ICs?


Is it something buried in Man’s psychological make
-
up that makes us constantly
strive to produce a better product?


The answer to the exi
stence of Moore’s Law depends on a variety of factors, some of
which, perhaps, are mentioned above.

Technological Advances


are they limited?

Intel

have announced that they have the technology to produce microprocessors
containing more than 400 million transistors, running at 10 gigahertz and operating at
less than one volt, in the next five to ten years. This is in line with Moore’s law


The trend t
oward smaller and smaller circuits will eventually lead to a situation where
the dimensions of the individual devices on the wafer approach atomic size. There is a
lower limit to the size that a component can be. If we extrapolate into the future we
can vi
sualise a time when each device is made with several key molecules and the
conductive tracks between devices are just one molecule thick. This would then be the
lower limit of micro
-
integration. It would not be possible to reduce the size any
further as we

cannot find anything smaller than individual atoms to build a circuit.


It is therefore inevitable that Moore’s Law will eventually conclude when we cannot
manufacture devices any smaller than atoms allow.


When the lower limit of device size is reached,

the speed that electricity travels within
the IC itself will also be a limiting factor. Electrical signals tend to propagate at
around 0.66c (2/3 the speed of light). The speed of propagation of any signal depends
on the medium through which the signal is

travelling, but the upper limit (i.e. c = 3 x
10
8

metres per second) cannot be exceeded within the laws that we have postulated to
form our universe.


Proposals for quantum storage (
quantum data registers

-

a faster, more efficient way
to store and retrieve data than the binary system we use today, because the rules of
quantum mechanics allow you to search many locations simultaneously),
light
operated transistors
, electro
-
optical polymers and more are showing new techniques
for achieving the ever higher performance demanded by industry and consumers.

Uni versity of Greenwich


EADS, DSE

Mark Clements

Page
5

11/2/2013

The Future of ICs

Whilst Moore acknowledged that his "law" won't hold forever,

he asserted that the
right technological approaches can delay "forever," extending the longevity of his
original prediction.
Read Intel’s ideas for ‘delaying forever
’.


Power cons
umption of ICs has fallen dramatically allowing complex devices such as
handheld computers to have very long operational times before requiring recharging.

As the circuits get smaller, their power consumption demands shrink too.


Moore’s Law does eventual
ly reach a lower limit due to the achievement of the
absolute minimum size of a particular component. Does this spell the end for Moore?
In the terms that he expressed his law, yes.


It is fairly obvious that manufacturers of ICs, such as Intel, will be re
searching ways
of making their ICs even more powerful once the limit to Moore’s law has been
reached. If we cannot increase the component density any further than the
manufacturers will attempt to make the existing devices operate at faster rates and
there
by increase the functionality of the existing devices.


New materials such as strained silicon and Silicon Germanium (SiGe) technology and
transistors that switch at speeds around THz (can switch on and off a trillion times per
second) are in development r
ight now.

Moore’s Law Version 2?


Just because there is a finite end to the size of a particular structure on a chip does not
mean that IC technology will stagnate. The manufacturers of ICs will strive to achieve
faster devices by using more diverse techn
ologies. If Moore’s Law remains counting
the number of devices per square millimetre, it will be of no use in predicting future
trends.


As with all technologies, revisions do occur to keep up with technological advances.
One suggestion that could be follo
wed is that perhaps the law needs to be revised in
the light of its impending demise.


Instead of looking at the number of devices per square millimetre of silicon wafer,
perhaps Moore’s Law version 2 could be developed. This new version would no
longer l
ook solely at the device count, but might take into account the performance of
the particular IC that has been manufactured. It could take into account the switching
time of the onboard devices and the capabilities of the devices themselves. For a
computer

microprocessor, it might look at the number of floating point operations that
could be performed per second. For a piece of memory, the Law might look at the
read/ write time of the device. Deference could be made to Moore version 1 by
looking at the aver
age performance of an IC per square millimetre.


Whatever the metric, or group of metrics, that could be employed to examine IC
performance, it is clear from the trend already in place that the performance of ICs
will continue to increase as long as there
are companies creating digital devices and
consumers exist to buy the products.

Uni versity of Greenwich


EADS, DSE

Mark Clements

Page
6

11/2/2013


Conclusion

Moore’s Law was formulated from observations made by counting the number of
devices on a particular area of a silicon wafer. Moore’s Law has remained true for
almos
t forty years. Extrapolation of the size of the devices on the silicon themselves
shows us that there is a finite limit beyond which it will become impossible to
manufacture a particular feature, such as a transistor. At this point Moore’s Law will
cease t
o have any meaning because the transistor count will not be able to increase.


Observations of Man’s ingenuity suggest that although the number of devices will not
be able to increase, the performance of the ICs themselves will constantly improve. A
revis
ed version of Moore’s Law could include IC performance figures and allow
future trends to be calculated past the demise of the current implementation of
Moore’s Law.


It remains for such a revision, titled here as Moore version 2, to be formulated and
test
ed against past and current trends.

Uni versity of Greenwich


EADS, DSE

Mark Clements

Page
7

11/2/2013

Appendix


The following data was used to produce the graph shown on page 1 of this document.


Device Name

Year of introduction

Transistor Count

4004

1971

2,250

8008

1972

2,500

8080

1974

5,000

8086

1978

29,000

286

1
982

120,000

386™ processor

ㄹ㠵

㈷㔬〰O

486™ DX processor

ㄹ㠹

ㄬㄸ〬〰1

mentium® 灲潣ess潲

ㄹ㤳

㌬㄰〬〰3

mentium ff 灲潣ess潲

ㄹ㤷

㜬㔰〬〰T

mentium fff 灲潣ess潲

ㄹ㤹

㈴ⰰ〰ⰰ〰

mentium 㐠灲潣ess潲

㈰〰

㐲ⰰ〰ⰰ〰


Uni versity of Greenwich


EADS, DSE

Mark Clements

Page
8

11/2/2013

References


http://www.intel.com/research/silicon/mooreslaw.htm

http://www.intel.com/labs/features/mi03031.htm