Re: FPGAs and Heat (Re: Paranoid Musings)

At 9:39 AM 8/1/96, Scott Schryvers wrote:

>This was about a year and a half ago.
>I can't remember the name of it,  but this chip fab industry mag was
>talking about how the NSA was obtaining out side help in fabricating what
>was at the time a type of ram that did processing off chip in parrallel.

This was a company in Bowie, Maryland, closely linked with the NSA and with
the "supercomputer centers." The idea of "processing in memory" has been
explored by various companies.

By the way, on the subject of using FPGAs for computers, here's a URL I
found that's interesting:

http://www.io.com/~guccione/HW_list.html

>Side note: Wired just recently talked about IRAM or Intelligent ram, and
>how it seems to be the future of high speed computation.

Side side note: I worked on Intel's "iRAM," standing for "intelligent RAM,"
in 1980-81. It found little market success. The idea of changing the
architecture of RAM bubbles up every few years, but has not yet succeeded
(except in some video-specific applications).

Cautionary Note: Bubble memories, laser pantography, integrated injection
logic, e-beam addressed memory, neural nets, Josephson junctions....

When you've watched the industry for enough years you'll learn to cast a
jaundiced eye on pronouncements that a technology is the Next Big Thing.
The above list--which covers only chips, not similar Next Big Things in
software--is a list of some of the things "Wired" would've hyped, had it
been published back then.

Most such announcements come out public relations departments at major
public labs, or from over-enthusiastic VCs. Or from claims made in papers
presented at the International Solid State Circuits Conference and similar
conferences. Reporters seeking stories then push the story.

The usual form of the press release goes something like this:

"The discovery of foobartronic switches may mean chips that are ten times
faster and one hundred times denser. Researchers say the foobartronic
revolution could reshape the entire industry..."

Few of the advances reported in "Wired" will ever see the light of day....
Some will, of course, but it's useful to remember that most of it is hype.

--Tim May


Boycott "Big Brother Inside" software!
We got computers, we're tapping phone lines, we know that that ain't allowed.
---------:---------:---------:---------:---------:---------:---------:----
Timothy C. May              | Crypto Anarchy: encryption, digital money,
tcmay@got.net  408-728-0152 | anonymous networks, digital pseudonyms, zero
W.A.S.T.E.: Corralitos, CA  | knowledge, reputations, information markets,
Licensed Ontologist         | black markets, collapse of governments.
"National borders aren't even speed bumps on the information superhighway."

Timothy C. May writes:

> This was a company in Bowie, Maryland, closely linked with the NSA and with
> the "supercomputer centers." The idea of "processing in memory" has been
> explored by various companies.

That's one of the things that killed Thinking Machines.  It turned out
that a standard supercomputer with PIM chips for memory could give the
same performance for less money.  The PIMs did the massively parallel
computation with the standard architecture redistributing data as needed
using high bandwidth scatter-gather operations and moves.

At the time Thinking Machines went under, Seymour Cray had a big contract
for Cray Computer to deliver a PIM Cray machine to the government, but he
missed some deadlines, got cancelled, and his company went down the tubes
as well.  Too bad, it would have been a nice box. 

BTW, I gave up trying to predict innovations after being dragged down to 
see an early version of Visi-Calc running on an Apple, and horribly 
insulting the developers with comments like "But why would anyone want 
to emulate a ledger sheet?" and "I hope you guys didn't spend a lot of
time on this."  After repeating such performances at startups like Lotus
and Infocom, I realized that predictive history was not one of my talents.

--
     Mike Duvos         $    PGP 2.6 Public Key available     $
     mpd@netcom.com     $    via Finger.                      $

> 
> Timothy C. May writes:
> 
> > This was a company in Bowie, Maryland, closely linked with the NSA and with
> > the "supercomputer centers." 
> 
> That's one of the things that killed Thinking Machines.  It turned out
> that a standard supercomputer with PIM chips for memory could give the
> same performance for less money.  

See:

http://cesdis.gsfc.nasa.gov/linux/beuwolf/beuwolf.html

Don is doing interesting things with less...

-- 
A host is a host from coast to coast.................wb8foz@nrk.com
& no one will talk to a host that's close........[v].(301) 56-LINUX
Unless the host (that isn't close).........................pob 1433
is busy, hung or dead....................................20915-1433

Mike Duvos wrote:

| That's one of the things that killed Thinking Machines.  It turned out
[...]
| At the time Thinking Machines went under, Seymour Cray had a big contract

	Just a nit, but Thinking Machines is still in business, and has
had their first few profitable quarters. www.think.com

Adam

-- 
"It is seldom that liberty of any kind is lost all at once."
					               -Hume

Adam Shostack wrote:

> | That's one of the things that killed Thinking Machines.  It turned out
> [...]
> | At the time Thinking Machines went under, Seymour Cray had a big contract
> 
> 	Just a nit, but Thinking Machines is still in business, and has
> had their first few profitable quarters. www.think.com

The current Thinking Machines is a software firm, and Daniel Hillis is no
longer amongst the top management.  The name lives on, but the business
of designing, building, and selling exotic supercomputers is kaput.

Not an uncommon story in the computer business.  Even Control Data 
Corporation still exists in a transmogrified form, although their
mainframe business went up in smoke ages ago. 

--
     Mike Duvos         $    PGP 2.6 Public Key available     $
     mpd@netcom.com     $    via Finger.                      $