Eugene.Leitl@lrz.uni-muenchen.de wrote:
> Adrian Tymes wrote:
> > * Come up with designs for a x86-compatible processor and motherboard -
> > not necessarily fast, not necessarily Pentium-compatible (386 should
> > be sufficient), but two layers at most - and print it on some plastic
>
> Why? You can build decent CPUs with ~10 k transistors.
Whatcha gonna *run* on 'em? This is a design for a business, and being
able to tap into existing infrastructure (in this case, software) helps
a lot with early rounds.
> > a la the printable PC research at MIT. Emphasis here is that the
>
> Why? Plastic transistors have their niches, but wearable CPUs are
> not it. Displays, possible.
See the next two.
> > circuit still works even with minor (~ 1 mm) holes and rips.
>
> If you're talking wearable hardware, I'll be very happy with a fannypack
> and a decent headset/headup. Sewing stuff into cloth is silly.
You would, but I'm not aiming for early adoptors. I'm aiming for the
mass market, which means aesthetics become part of the functionality:
they look wierd or are at all uncomfortable, and they're as good as
busted right off the manufacturing line. You can hide a lot in a vest,
and make it look slick in many social environments.
> > Possibly use insane amounts of parallel processing (since "chip" real
> > estate is relatively cheap) to help mitigate speed issues brought on
>
> Chip real estate is not cheap, not in the world I woke up in today.
Not on silicon, nor on most tiny chips. But when you've got area on
the order of a person's whole back to work with...
> > by large feature size (at 100 MHz, even light can only travel 3
> > meters per cycle, counting all the circuit traces would have to go
> > through; if each feature is, say, 2 mm wide, then that's at most
> > 1500 gates even if one could put them right next to each other, and
> > that's not even accounting for memory latency).
>
> If your transistors are made from plastic, relativistic latency
> will be the least of your problems.
Yeah, but speed will be a problem and must be dealt with nonetheless.
This archive was generated by hypermail 2b30 : Mon May 28 2001 - 09:50:39 MDT