Supercomputers Are Still Fast, but Less Super
Anne & Lynn Wheeler
http://bits.blogs.nytimes.com/2009/11/18/supercomputers-are-still-fast-but-less-super/
from above:
The supercomputing world was long dominated by systems that required
specialized chips, memory systems and networking technology. But about
10 years ago, researchers realized they could link thousands of cheaper
machines running on mainstream chips and achieve pretty solid
performance.
... snip ...
some recent threads on the subject:
http://www.garlic.com/~lynn/2009h.html#1 z/Journal Does it Again
http://www.garlic.com/~lynn/2009j.html#37 A Complete History Of Mainframe Computing
http://www.garlic.com/~lynn/2009k.html#50 A Complete History Of Mainframe Computing
http://www.garlic.com/~lynn/2009o.html#57 U.S. begins inquiry of IBM in mainframe market
http://www.garlic.com/~lynn/2009o.html#81 big iron mainframe vs. x86 servers
http://www.garlic.com/~lynn/2009p.html#54 big iron mainframe vs. x86 servers
http://www.garlic.com/~lynn/2009p.html#55 MasPar compiler and simulator
referencing this old news snippets from '92
http://www.garlic.com/~lynn/2001n.html#6000clusters1
http://www.garlic.com/~lynn/2001n.html#6000clusters2
--
40+yrs virtualization experience (since Jan68), online at home since Mar1970
Quadibloc
specialized chips for supercomputing, such as were found in the Cray
X-1 and the NEC SX-6, are no longer common. However, as more and more
area becomes available on silicon, there apparently is a chance that
Intel will eventually add vector supercomputer-like instructions (that
is, based on the Cray-1 style of vectors, not the TX-2 and AN/FSQ-32
style of vectors) to a forthcoming iteration of SSE.
The article doesn't note that Blue Gene is based on PowerPC chips,
which are still conventional microprocessor chips rather than
specialized supercomputer chips.
John Savard
Joe Pfeiffer
> I have to admit, I find it somewhat dismaying that architecturally-
> specialized chips for supercomputing, such as were found in the Cray
> X-1 and the NEC SX-6, are no longer common. However, as more and more
> area becomes available on silicon, there apparently is a chance that
> Intel will eventually add vector supercomputer-like instructions (that
> is, based on the Cray-1 style of vectors, not the TX-2 and AN/FSQ-32
> style of vectors) to a forthcoming iteration of SSE.
That would be really cool...
> The article doesn't note that Blue Gene is based on PowerPC chips,
> which are still conventional microprocessor chips rather than
> specialized supercomputer chips.
>
> John Savard
--
As we enjoy great advantages from the inventions of others, we should
be glad of an opportunity to serve others by any invention of ours;
and this we should do freely and generously. (Benjamin Franklin)
Quadibloc
> Quadibloc <jsav...@ecn.ab.ca> writes:
> > I have to admit, I find it somewhat dismaying that architecturally-
> > specialized chips for supercomputing, such as were found in the Cray
> > X-1 and the NEC SX-6, are no longer common. However, as more and more
> > area becomes available on silicon, there apparently is a chance that
> > Intel will eventually add vector supercomputer-like instructions (that
> > is, based on the Cray-1 style of vectors, not the TX-2 and AN/FSQ-32
> > style of vectors) to a forthcoming iteration of SSE.
>
> That would be really cool...
It indeed would be, but perhaps I was misreading the news item due to
wishful thinking on my part. Rather than a future general Intel chip,
whatever feature was discussed might only show up in custom x86 chips
Intel was to make for a Japanese supercomputer maker.
John Savard