[HH] The Death of Moore's Law Will Spur Innovation

[Bill Bogstad]

On Fri, Apr 10, 2015 at 9:37 PM, Tom Metro <tmetro+hhacking at gmail.com> wrote:
> Here's an opinion piece by Andrew "bunnie" Huang (hardware lead at
> Chumby, and I believe the guy I previously posted about who was building
> an open hardware laptop design).
>
> The Death of Moore's Law Will Spur Innovation
> http://spectrum.ieee.org/semiconductors/design/the-death-of-moores-law-will-spur-innovation
>...
>
> This all sounds great for those rooting for open hardware, but all of
> this flows from the starting assumption that Moore's Law will run out of
> gas. The problem with that assumption, even though it is supported by
> the laws of physics, is that there will be many highly motivated
> organizations with deep pockets that will seek to redefine the problem.

Actually, I don't think it is that startling at all.   Whether it's
multi-core, VLIW, or adding more and more specialized instructions
(SSE, SSE2, SSE3) to optimize particular applications; it seems that
deep pockets have been attempting to redefine the problem for a long
time now.   This does basically nothing for legacy apps and frequently
not that much for most newly written apps either.    I'm aware of only
two broad areas which might overcome the fast approaching limits of
physics:

1. Switch from silicon to some other substance.   I periodically read
articles about carbon nanotubes or other alternatives.   There seems a
fair amount of money being poured into this and it's not yet clear
that these efforts won't bear fruit.   But then again maybe not.

2. Move more towards 3D features in semiconductors.   Again it seems
lots of money is being spent here.  My impression is that it can be
made to work, but it is likely to significantly increase manufacturing
costs.   If you REALLY need single package improvements in density
this will be doable but it seems like it will be hard to justify for
mass market products.

Another issue which neither of you mention is the increasing cost of
the plants where these chips are produced.   As I understand it, even
Intel finds it expensive to keep pushing for the next incremental
shrink in chip technology.   If integrated design/manufacturing firms
become just too risky (what if a chip design is late? do you let your
multi-billion dollar factory sit idle?), then I would expect to see
even more contract chip manufacturing plants to appear.  This may make
it possible for smaller chip design firms to get access to current
manufacturing technology.   If that happens, the next step would be
for design "building blocks" to either be made available by the chip
foundries or as open designs.   Smaller chip design firms will also
probably find it easier to use standardized interfaces/design
components as much as possible in order to reduce their cost/time to
market.   This seems to me to be another influence that will push the
industry in the direction that Andrew Huang suggests.

Bill Bogstad