[mdlug] Parallella: A Supercomputer For Everyone by Adapteva — Kickstarter

[Aaron Kulkis]

Garry Stahl wrote:
> On 10/10/2012 09:12 AM, Adam Tauno Williams wrote:
>> On Wed, 2012-10-10 at 06:04 -0700, David Lane wrote:
>>> I have looked into high performance computing the question is:
>>> What application do you have that needs supercomputing?
>> +1 This is 'cool', but 99.44% of users need a supercomputer as badly as
>> they need a giant pink inflatable elephant.
>>
>> But that other 0.56% need one yesterday.
>>
>> A supercomputer is only useful for problems that can be decomposed in
>> certain ways [or is worth figuring out how to decompose].  Otherwise an
>> old-fashioned just-faster-computer is better.
>>
>
> As the gent pointed out "just faster" is hitting physical limits.
>
>

Yes.  In the 1960's, Richard Feynman gave a speech encouraging
scientists and engineers to make things smaller, titled, "There's
plenty of room at the bottom."  This is considered one of the
key messages in waking up the Sci/Eng community to miniaturization.

These days, transistor sizes are now down to a couple hundred
atoms.  And unfortunately, smaller element sizes means using
higher frequency energies.... which in turns causes less
horizontal displacement (if the wafer horizontal), but
more vertical displacement and variation of dopants within
the substrate material (i.e. the "silicon wafer") or what
have you.

Basically, we're already running into aspects of the
Heisenberg Uncertainty Principle.  And we're not even
close to transtors and diodes with sizes on the order
of 10 atoms or so (although IBM did make a transistor
of something like 5 atoms.... there's no way to mass
produce something like that -- it was a publicicity
stunt, pure and simple.)