[rescue] Perverse Question
Francisco Javier Mesa-Martinez
lefa at ucsc.edu
Sun Jun 15 18:30:04 CDT 2003
> From waht I'm given to understand, one hell of a lot of the power
> is consumed (and, correspondingly, heat produced) just by the clock
> circuitry. These boys could learn something from the async-logic
> "of old".
Sometimes comments like these get me, us design "boys" are perfectly aware
of that, the whole async business being the holy savior of
computer architecture is bogus. The tradeoff is that potentially async
logic (whatever that is really) has a limited gain in power dissipation, while at
the same time it increases the design complexity a big deal designing a
fully asynch processor is a nightmare. Switching network theories are
better understood that fully async networks. The proposers of the
technology blissfully ignore those details, these people just think that
clocked designs came out of the blue. There is a reason!
BTW there is plenty of asynch logic in modern systems. Most of the logic
inside each pipe stage is actually async... Think of a pipeline as the
glue to put together a bunch of async pieces. I am just amazed how many
times I have read people inveting the wheel for the ntime claiming the
revolution is here :).
So the question is there: Do you design a superduper async system that
takes 5 years to get out of the door, or you bet for a traditional design
which is cheaper and is out of the door in 1 year? Granted is not as
superduper by a factor of 10%. The problem is that by the time you release
the superduper system, the old fart system has gone through 5 generations
and you are only competive with the 1st generation, i.e. 4 year old
technology... This is what everyone has had to deal with: the steam roller
that is CMOS.
> > The poor (and *ancient*...people say VAX is old? It's newer than
> > x86!) architecture has backed them into a very dark corner in which the
> > only performance gains can be had by increasing the clock rate...which
> > is nearing the end of its rope.
They have been saying this for over 20 years. I am skeptical as usual,
Intel has the best production technology period and they know it... thus
they are taking advantage of it. Is elegant, not IMHO... but then again
who does get to define what elegant is?
CMOS was not supposed to go past 40MHz at some point. Thus the whole rush
for ECL, CMOS was supposed to get too hot after 100Mhz then the whole rush
for GaAs. Maybe third time is a charm for CMOS but I really doubt it. I do
not really like it though :).
> > Thermal migration of materials (diffusion) within the chip is also an
> > issue, but PC hardware isn't [generally] designed for longevity so
> > that's not really a big deal.
> Give it five years running at a core temperature that must approach
> the boilling point of water (if not exceed it)....
Many big iron was running under those conditions for eons, whoever the
life cycle of most moder FeeCees is under 2 years.
Again tradeoffs, design a better PC which is more expensive, or just get a
cheaper machine out of the door asap. Pure Capitalism is way too
inefficient sometimes because it tends to get stuck in local minimae way
too often :).
More information about the rescue