Information Technology Reference
In-Depth Information
Vemuru:
I have a comment on the Moore's Law. Moore's Law is great not
because you could integrate things, but price goes down exponentially. It is a
good economic model for things to grow. If we can deliver QCA - say, processes
to make a few cells - down the line if you could make a bigger area with the
same cost then we will have Moore's Law in reverse order. I think that's the
path we should choose.
Porod:
I would like to echo the second part. There are obviously different ver-
sions of Moore's Law. Because we like technology we think about sort of the
ones we relate to the technology feature size and number of transistors and so
on. But what really drives things is economics. People don't spend money on a
new computer because they are so proud of the technology that goes into it -
because they have 22nm feature size transistors in there - no one cares about
that. People care about performance. So as long as you can deliver performance,
the way you do it does not matter and Moore's Law continues.
In recent years, as you know, the industry has actually backed off of raw
transistor performance in favor of parallelism, and that has kept performance
improving. So 2 cores, 4 cores, 8 cores and so on. In principle, you could envision
going to million cores if you can place a billion transistors on a chip. One way of
arranging a billion transistors is in an array of a thousand by thousand process-
ing elements - an array of a million processing elements - each with thousand
transistors (which is the order of complexity of one of the early microprocessors).
So I think there is a long way to go for increased performance, if we just exploit
the parallelism. I think this is a continuation of Moore's Law, if you will. It has
moved from raw technology to architecture, but consumers do not care.
Anderson:
Moore's Law by other means. Thank you all, and thanks to the
panel.
Acknowledgments.
We thank all of the FCN panelists and participants for their con-
tributions, USF graduate students Srinath Rajaram, Ravi Panchumarthy and Dinuka
Karunaratne for their work recording the panel discussion, and Katherine Anderson
for transcription assistance.