koworld wrote:
a) predicatble and b) broadly dependent on Moore's law.
Several thousand engineers would disagree.
a) The predictability of future IC performance is going right out the
window. Clock speed isn't scaling the way it used to, the increase in
transistors isn't leading to a proportional increase in peformance,
heat is becoming a very big problem at the high end.
b) Moore's Law is not a 'law' but an observation, which in itself only held
broadly true because organsitations like SEMATECH incorporated the
predictions into their reasearch and development plans. i.e. It became
a self-fullfilling prophecy. Moore simply observed that the number of
transistors in a low-cost IC was doubling roughly every 12 months
which he later revised to two years. It's held true for many years because
in the early days it was easy to improve IC production, and laterly it
became the target to reach at all costs, now though nature is getting in
the way of future progress. You can only send a signal so fast, more
and more of the current you put in leaks away from the circuit and produces
heat, you can't scale transistors down indefinitely (quantum effects start kicking
in). On top of the physical effects there are associated circuit problems, memory
is realtively slower and slower to computation, circuits have become more
complicated to overcome physical effects (wire delay, clock skew, leakage),
high performance is requiring increasingly complex circuits to avoid bottle-necks
and stalling.
Any how the long and the short of it is that the kind of annual improvements that
people have come to expect are no longer panning out and other directions are
being sought ASAP. One of which would be Cell for the PS3... which I may
expand upon later with my 3rd/4th hand knowledge of it.
fluffiest 
forum 
in town 
Ermm ... not wishing to pry but what do you call "old"? 
