How fast (ie: clockspeed in Hz) can a single processor in a computer be? The highest I've read is 4 GHz. What's the limit?
It is difficult to speak of an absolute frequency limit, since there are many
different limitations on clock speed, some of which are technical, and some are
One important assumption in circuit design is that all circuit elements are
'lumped'. This means that signal propagation time from one element to the other is
negligible. 'Negligible' means that the time it takes for the signal produced at
one point on the circuit to propagate to the rest of the circuit is (very) small
compared to the times involved in circuit operation.
For all practical purposes, electrical signals travel at the speed of light.
Let us take an example: Assume a processor which works at 1GHz. This means one
billion clock cycles per second. This also means one clock cycle takes one
billionth of a second, or a nanosecond. Light travels about 30cm (about a foot) in
a nanosecond. So, the size of circuitry involved at such clock speeds better be
much less than (at least 1/10 of) 30cm. So, your maximum circuit size is 3cm.
Taking into account that the actual CPU core size is less than 1cm a side, we are
still in safe waters.
Remember that this was for 1 GHz. If the clock speed is increased to 100GHz, a
cycle will be 0.01 nanoseconds, and signals will only propagate 3mm in this time.
So, your CPU core will ideally need to be about 0.3mm in size. It is quite hard to
cram a CPU core into such a small space. So, we're still in safe waters, but
somewhere between now and 100GHz, we're going to hit this physical barrier.
What happens when this size limitation is violated? What happens is that
certain parts of your circuit are in the 'current' cycle, and other parts are in
the 'previous' cycle. It gets really hard to handle such distributed systems. Then
in design, you'll have to handle delays in signal propagation, circuit shape starts
playing a strong role in design. In fact, it becomes more like separate processors
than a single processor.
Then there are other technical problems, like overheating, circuit design,
perhaps having stable clocks at such high frequencies and such.
But I believe none are as serious as the physical limitation of clock speed vs.
Yasar Safkan, Ph.D., Sofware Engineer, Noktalar A.S., Istanbul, Turkey
Present day microprocessors (like the Pentiums) are built with the
complementary-MOS (CMOS) technology. How fast a processor can be clocked finally
depends on how small a switching time each MOS transistor can have. The switching
time of a MOS Field Effect Transistor (MOSFET) goes down with its size. So, as long as
you can make transistors smaller, you can get faster transistors and hence faster
processors. This reduction in the size is termed as 'CMOS Scaling' and has been
more or less following the celebrated Moore's Law.
The smallest feature size to be found in a processor today is some 150 nanometers
(this is the so called 'channel length' in a MOSFET). You can go smaller than that,
but with ever increasing difficulty. Conventional transistors may still work
down to about 50 nanometers channel length, giving your processor clock speeds
around 20 GHz or so, but beyond that, its anybody's guess. At such small lengths, a
transistor isn't really well behaved, so one can't use it as one has been. Quantum
effects reign supreme.
One can actually try to put some of these quantum effects to good use and make
devices (like transistors) based on them. Single electron transistors (SETs) which
have shown work at room temperatures hold promise in this regard. It may be
possible one day to make integrated cicuits out of them. That would give processors
speeds of as much as a terahertz. But one can be sure of not seeing any of these
things in the market in the next 20 years at least. Till then, its old CMOS,
inching towards higher speeds and smaller dimensions, saturating its potential in
Palash Bharadwaj, Electrical Engineering Undergraduate, IIT Bombay
'An expert is someone who knows some of the worst mistakes that can be made in his subject and how to avoid them.'