[ExI] Two Paths to Further Shrink Computer Chips

[samantha]

 On 9/2/10 3:47 PM, John Grigg wrote:
> Tomasz Rola wrote:
> This sounds nice, gives hopes that one day building a 1000-core cpu will
> be trivial (and I feel pervert pleasure thinking I would one day buy it
> used from internet for the cost of pocket radio).
> My question to everyone here...  How many years until we have a
> 1000-core cpu (or something equivalent) as a household desktop
> computer?  My optimistic estimate is within ten years or by 2020.  And
> what exactly would the number crunching/calculating power of a
> 1000-core cpu be?

About 5 years would be my guess.  Sooner in GPUs.  Which are probably
part of the development path to such but made fit for more general
computation.
> I thought at first such a machine would simply be 500 times more
> powerful than let's say a current dual processor computer, but
> considering each individual processor may easily be a hundred times as
> powerful as a cpu now, I would think this hypothetical machine would
> literally be thousands of times more powerful than an average pc from
> today.  Oh, and what about price?

I doubt you will get both at the same time in that timeframe.  Massively
more cores seems to be the current direction rather than much faster
clock rate.  I think it will take longer to get things like hyperfast
graphene transistors out of the lab and into consumer chips.   I do
think you will see massive static memory chips for very cheap in about
about the same time frame and a couple of orders of magnitude faster. 

> I'm looking forward to being gee-wowed!!!  : )
>
Most of the software you run will probably bloat to match.  :P

- s