RE: filmscanners: That's some overclocking

[derek_c@cix.co.uk (Derek Clarke)]

CPUs _multiply_ the external clock these days as the PCB can't carry 
nearly so high a rate as the chip can without a lot more expensive 
engineering.

The issue with the speed of electrons is a valid one, which is why CPU 
architectures are headed more towards getting more out of every clock 
cycle than too much of an increase in clock speed.

Gallium arsenide is tough to process, so I would be more inclined to 
expect a shift to optical circuits in the future. They at least operate at 
the speed of light rather than the somewhat slower rate of electrons, and 
the switching time of an optical gate can be very short indeed.

Eventually quantum computers may be able to do the whole job in one clock 
cycle, in which case the clock speed will be less important :-)

mgduncan@esper.com (Mike Duncan) wrote:

> >I think something is a little overclocked. <BG>  The speed of
> >light is 2.99 times 10E10  (that is 3 followed by 10 zeros) cm
> >per second.  One of the outcomes of Einstein's theory of
> >relativity is nothing can travel faster then the speed of light.
> >
> >I don't know the size of the actual chip, but if the chip was 1
> >cm long and presumably an electrical signal would have to travel
> >the length of the chip sometime, then in a single cycle that fast
> >electron would travel 1 cm.  That would be an average speed of 7
> >times E10 cm per second or more then twice the speed of light.
> >Note I said average speed.  Since the electron must start and
> >stop the actual top speed would need to be even faster.
> >
> >In fairness I beleive chips are  smaller then 1 cm (but larger
> >then 0.1 cm), so my little argement is not valid; however, today
> >distances and the time to travel those distances are a
> >significant part of the limitation for chips.  So I feel with
> >some confidence the 70 gHZ number is not possible.  I would
> >personally be amazed at a number of 7 gHZ with the currently
> >available chip manufacturing processes -- using Xrays to layout
> >the grid might make  that possible.
> 
> You are correct, propagations are one limit to clock rates.  Some CPU's
> actually divide the clock frequency to lower rates internally.
> 
> Mike Duncan
> 
> 
>