Re: How Fast Could You Get A Superconducor CPU To Go?

Hi Nathan!

It is indeed true that in recent years current CPU technology has been
going to lower voltages to reduce heating in microprocessors and thus
allowing more current to flow inside. This is actually not the only way to
reduce heating, but the basic thinking is thus...

We know (hopefully) that the power dissipated by an resistor is
P = VI
(power dissipated) = (potential difference) x (current flowing)

so if you drop the working voltage of a CPU then you can dissipate
less heat for the same current.

Your suggestion is then that we use a superconducting device and high
voltages, but if we substitute back for Ohm's law in the power
equation above we get P = V x V / R so
if the voltage gets bigger and the resistance drops the power
dissipated goes to infinity!!! That is not good. 

In fact, if we were to use a superconductor then we wouldn't need to
have high voltages since we could make the same amount of current flow
for less volts. However, superconductors (at the moment) require
temperatures below 100 Kelvin (-173 degrees Celcius) and are therefore a
little impractical to use for home or office computers. It is also
difficult to make logic gates out of superconductors. These materials are
really not suited to making microprocessors out of in their current state.

However, silicon technology can still go a long way... The original
question was about speed so let us first ask ourselves what limits the
speed of an electronic switch? 

You were right in saying that the resistance of the device contributes
to the speed limit. Resistance causes heating and the last thing that
we want is a CPU that cooks itself to death. However, reducing the
current flowing along with the resistance is a better plan. Recall P=I*I*R
too.

The second, and equally important thing, is capacitance. An electronic
switch works like a valve to turn on and off current flow. The current flow
can be very very small as long as it is measureable by other devices, but
the capacitance of the switch controls how fast it can be switched on and
off. The capacitance of the device can be controlled by making smaller and
smaller transistors and using different transistor technology on the chip.
This is where the effort is made in the industry.

Intel have some nice pages on this... probably the most important of
which is Moore's law which is related to speed and the number of
transistors on a chip: http://www.intel.com/research/silicon/mooreslaw.htm

Then check out their plans for Terahertz (1000 GHz) transistors: 
http://www.intel.com/research/silicon/TeraHertz2.pdf

There was also an article this month in the IEEEs magazine "Spectrum"
which talks about the future of microprocessors which is quite
interesting and says some interesting things about the current state of the
art:
 http://www.spectrum.ieee.org/WEBONLY/publicfeature/apr02/mlaw.html

I am not sure I believe all of what is suggested in this article though.

Here I have only just scratched the surface of the problems in making fast
microprocessors. There are many other issues and some related to what
happens when the transistors on a CPU are the size of the silicon atoms
themselves, but this is another story.

Phil.

---

Dr. Phil Marsden
--------------------------------------------------------------------------
Department of Microelectronics and Information Technology (IMIT)
Laboratory of Optics, Photonics and Quantum Electronics (OPQ)
Royal Institute of Technology (KTH) Stockholm
--------------------------------------------------------------------------