🚀 go-pugleaf

RetroBBS NetNews Server

Inspired by RockSolid Light RIP Retro Guy

7 total messages Started by Siamak Hassanzadeh Mon, 27 Jul 1992 17:55
Fujitsu Supercomputer Statement of Direction
#17
Author: Siamak Hassanzadeh
Date: Mon, 27 Jul 1992 17:55
136 lines
7317 bytes
        Fujitsu  Supercomputer  Statement  of  Direction

           FUJITSU UNVEILS ITS SUPERCOMPUTING STRATEGY


(Washington, D.C. July 21, 1992) -- Fujitsu Limited today
outlined its supercomputer development and marketing strategy at
the 6th ACM International Conference on Supercomputing.  As part
of this announcement, Fujitsu disclosed its development of a
powerful parallel supercomputer capable of peak processing speeds
of more than 300 gigaFLOPS (300 billion floating point operations
per second) using only a few hundred processors -- an unprecedented
performance level.

     Central to Fujitsu's product strategy is a recognition that
existing approaches to supercomputing, i.e. vector processing
and parallel processing, each have their respective merits.
Furthermore, Fujitsu has found that combining elements of both
architectures, that is, coupling Fujitsu's expertise in vector
processing technology with the growing promise of parallel
processing, yields a third and even more powerful approach to
supercomputing.  This newly developed architecture provides a
critical third element to Fujitsu's three-pronged supercomputer
product strategy.

     Known for its VP and VPX series of vector supercomputers,
Fujitsu has acquired a great deal of knowledge and experience in
vector processing.  Fujitsu has also gained expertise with its
AP1000 system, a prototype highly parallel computing system based
on scalar processing technology.  The AP1000 has yielded important
research results.  Fujitsu's scientists have learned that parallel
supercomputing requires a radically different approach to programming.

     Equally important, computational case studies have revealed
that conventional vector processing, vector parallel processing,
and parallel scalar processing each excel at different applications.
In some cases, computational efficiency is improved through the
parallelization and/or vectorization of an application.  In other
cases, the complexity of parallelizing software code makes
conventional vector processing the best option.

     Having experimented with all three architectures,
Fujitsu has developed a clear understanding of how to best serve
the needs of supercomputer users.  Fujitsu's research, including
joint studies with users, has demonstrated that the third
approach, namely vector parallel processing, is very efficient
over a wide range of applications.  Nevertheless, conventional
vector supercomputers continue to provide superior performance
for many computational applications.  Therefore, Fujitsu is
committed to advancing the state of vector supercomputing as well
as parallel computing and plans to release succeeding generations
of its VP and VPX Series.

     With this announcement Fujitsu reveals the development of a
new generation of ultra-high performance systems based upon a
parallel arrangement of vector processors.

     Fujitsu's highly parallel approach maximizes single
processing unit performance, packing gigaFLOPS performance into 
each custom processor used in the parallel configuration.  Each
individual processor, which is the size of a large briefcase,
delivers supercomputer performance in and of itself.

     With hundreds instead of thousands of processors, Fujitsu's
strategic technology direction successfully addresses one of the
most critical issues in the parallel computing arena:  creating
teraFLOPS power without the restrictive programming environment
of today's massively parallel systems.  This approach combines
the optimal performance of parallel computing with the
reliability and ease of programming of vector processing.

     Based on this highly vector parallel approach, Fujitsu's new
supercomputer system, when introduced later this year, will offer
the worldwide scientific and engineering community state-of-the-
art computational capability to solve "grand challenge" issues
such as global warming and ozone depletion as well as industrial
and commercial computational problems such as large scale fluid
dynamics, anti-cancer drug development, seismic data processing,
and weather forecasting, for example.

     The key to Fujitsu's success in developing this vector
parallel system is its far-reaching experience in designing and
building conventional vector processors.  The advantages of
vector processing architecture are well proven, and by marrying 
this technology with the principles of parallel processing,
Fujitsu has developed a practical, ultra-high speed computer.
Fujitsu's approach has been to unite vector processors in a
parallel architecture, while others have attempted to add vector
features to existing parallel systems.

     In approaching the design of this system, Fujitsu did not 
rely on commodity, "off-the-shelf" processors, as do many of
today's massively parallel systems.  Each processor or processing
element includes a vector and a scalar unit along with a main
memory unit configured in a shared architecture.  Fujitsu's
unique distributed memory architecture affords maximum
flexibility by allowing the configuration of memory and
processing elements to match actual data processing requirements.

     "We plan to offer out customers the security and certainty
of today's vector supercomputing environment, a universal UNIX
operating system and commonly used FORTRAN programming language.
The big benefit to customers will be that millions of dollars of
software development assets can be preserved, without substantial
modification, as customers move to parallel capabilities and
teraFLOPS performance.  We also expect to offer the 10 teraFLOPS
performance level before the end of this decade," said
Tadashi Sekizawa, Fujitsu's President.

     Fujitsu Limited is the world's second largest computer 
manufacturer and the largest in Japan.  The company entered the
supercomputer market in 1977 with its FACOM 230-75 APU and today
offers its VP and VPX Series, a complete line of vector processor-
based supercomputers.  Today, Fujitsu ranks second worldwide
in the number of supercomputer installations with more than 160
around the globe.

     With more than $25 billion in revenues for the fiscal year
ended March 31, 1992, Fujitsu invests more than 10 percent of its
revenues annually on research and development to advance its
three core technology areas of focus:  electronic devices,
computers and communication systems.
 
                              # # #

UNIX is a registered trademark of UNIX System Laboratories, Inc.




-- 
      ***************************************************
      *    Siamak  Hassanzadeh        siamak@fai.com    * 
      *    Fujitsu America, Inc.      (408) 456-7308    *
      *    3055 Orchard Dr., San Jose, CA 95134-2022    *
Re: Fujitsu Supercomputer Statement of Direction
#19
Author: Dik T. Winter
Date: Tue, 28 Jul 1992 02:08
18 lines
1346 bytes
In article <Bs274H.BA5@news.fai.com> siamak@ossi.com (Siamak Hassanzadeh) writes:
 >            FUJITSU UNVEILS ITS SUPERCOMPUTING STRATEGY
...
Sometime ago somebody wondered about the amount of press-releases that
were posted from about the new Cray.  Well, this is the third time I
read this press-release.  Can't FAI coordinate their press-releases?
I found one back in comp.parallel (same text, different poster), and
there has been a third one, but I cannot find it; probably it was over
4 days ago.  I think that posting press-releases is not bad in itself,
what makes it bad is posting the same press-release multiple times to
either the same or to different newsgroups.  The last time I objected
to this was with Alliants campus effort.  Look where that got them ;-).

Peeved, dik
-- 
dik t. winter, cwi, kruislaan 413, 1098 sj  amsterdam, nederland
home: bovenover 215, 1025 jn  amsterdam, nederland
dik@cwi.nl
Re: Fujitsu Supercomputer Statement of Direction
#20
Author: Jason W Solinsky
Date: Tue, 28 Jul 1992 12:58
11 lines
1008 bytes
I have seen alot of press releases that were low on content, but after finding
this one for the third time, I have decided that it definitely takes the cake.
Without getting rid of any facts about the computer or directions of the company
this should be rewritten as:

  After experimenting with several parallel technologies, Fujitsu has build a
  300 GFLOP machine with hundreds of processors.

The end.

Jason W. Solinsky
Re: Fujitsu Supercomputer Statement of Direction
#21
Author: Daan Sandee
Date: Tue, 28 Jul 1992 13:46
18 lines
1373 bytes
In article <1992Jul28.125822.17675@athena.mit.edu>, solman@athena.mit.edu (Jason W Solinsky) writes:
|> Without getting rid of any facts about the computer or directions of the company
|> this should be rewritten as:
|> 
|>   After experimenting with several parallel technologies, Fujitsu has build a
|>   300 GFLOP machine with hundreds of processors.
|> 
|> The end.

Apparently the hundreds of lines of marketeers' language did its job. They
have fooled you.
Actually, they say they are *developing* such a system. They didn't say
they *have built* it yet. Scheduled for later this year, but they
don't say whether that is the 300-GFLOPS system or a smaller one.

Daan Sandee                                           sandee@think.com
Thinking Machines Corporation
Cambridge, Mass 02142                                 (617) 234-5044
Re: Fujitsu Supercomputer Statement of Direction
#22
Author: Jason W Solinsky
Date: Tue, 28 Jul 1992 15:25
22 lines
1792 bytes
In article <153j3gINN86a@early-bird.think.com>, sandee@Think.COM (Daan Sandee) writes:
|> In article <1992Jul28.125822.17675@athena.mit.edu>, solman@athena.mit.edu (Jason W Solinsky) writes:
|> |> Without getting rid of any facts about the computer or directions of the company
|> |> this should be rewritten as:
|> |> 
|> |>   After experimenting with several parallel technologies, Fujitsu has build a
|> |>   300 GFLOP machine with hundreds of processors.
|> |> 
|> |> The end.
|> 
|> Apparently the hundreds of lines of marketeers' language did its job. They
|> have fooled you.
|> Actually, they say they are *developing* such a system. They didn't say
|> they *have built* it yet. Scheduled for later this year, but they
|> don't say whether that is the 300-GFLOPS system or a smaller one.

Actually I didn't pay attention to that part because I know the USA will
make sure that its laws (both written and in the works) prevent its profitable
sale in the US. I'm not unhappy about that, BTW. In fact I rather like it, but
its the primary consideration when talking about a Japanese supercomputer.

Jason W. Solinsky
Re: Fujitsu and America
#24
Author: Greg Wilson
Date: Wed, 29 Jul 1992 07:19
15 lines
1421 bytes
In article <1992Jul28.152513.27457@athena.mit.edu>, solman@athena.mit.edu (Jason W Solinsky) writes:
> ....I know the USA will
> make sure that its laws (both written and in the works) prevent its profitable
> sale in the US. I'm not unhappy about that, BTW. In fact I rather like it...
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> Jason W. Solinsky

Aah, America, the home of free enterprise ... 

 ========================================================================
  Gregory V. Wilson                    ||  For in much wisdom is much
  Technical Coordinator                ||  grief; and he that increaseth
  Edinburgh Parallel Computing Centre  ||  knowledge, increaseth sorrow.
  gvw@epcc.edinburgh.ac.uk             ||               - Ecclesiastes
 ========================================================================
CFD code required
#40
Author: Armando Fortuna
Date: Mon, 03 Aug 1992 11:41
22 lines
1086 bytes
Hi netters,
I'm looking for a CFD code (public domain, or any that can be obtained for
a licence extension) that can handle:

_ 3D axisimetrical domains,
- compressible/incompressible flows,
- some heat transport (not necessary, but it would be usefull)
_ (volume) flow ranges between 0 and 2 liters/second
_ Reynolds numbers less than 2000

I'm working with human expiratory flows.

Any literature that can be suggested would also be MOST HELPFULL.

Any help would be very much appreciated.
Thanks everyone,

Armando Fortuna
fortuna@cs.man.ac.uk
University of Manchester
Department of Computer Science
Manchester, UK
Thread Navigation

This is a paginated view of messages in the thread with full content displayed inline.

Messages are displayed in chronological order, with the original post highlighted in green.

Use pagination controls to navigate through all messages in large threads.

Back to All Threads