Contents Issue 10 June 1999
|
|
| Europe keeps its position in the TOP500
|
|
This morning, the new TOP500 list of world most powerful supercomputers was released at the Mannheim Supercomputer 99 Seminar. With 134 machines, out of 500, Europe has improved a little bit over the 120 half a year ago. The USA are still dominating the list. Japan is falling back.
Also for the first time since two years, a European designed and manufactured machine made it to the TOP500. In Europe, Germany is still the largest user of supercomputers.
|
| ASCI Red still on first position in TOP500
|
|
|
Surprisingly, not one of the new ASCI Blue machines, did get
on the first position of the TOP500 list,
but the upgraded Intel ASCI Red system at Sandia National Laboratory with a performance of over 2.1 Tflop/s using 9472 Pentium processors remains the number one computer.
The ASCI Blue Mountain system at Los Alamos National Laboratory is number two with 1.6 Tflop/s.
|
| New record of participants to the 14th Mannheim Supercomputer Conference
|
|
|
For this year's edition
of the Mannheim Supercomputer Conference, which is already the fourteenth, a new record of 225 participants was registered.
In his opening speech, organizer Prof. Dr. Hans Meuer
proudly presented the content
of the three-day
event, taking place from June 10th till 12th.
The main theme of the first day, which traditionally
is Europe Day, constitutes " HPC in the next ten years".
The second day is dedicated to several highly attractive
sessions covering the universe of applications and tools and the Teraflops activities in the United States.
Short 15 minutes presentations have been planned to highlight the ongoing work of the HPC manufacturers. Saturday is devoted to SuParCup Day with the session on High Speed Cluster Computing with four talks on this important and evolving field. The big final of the Supercomputer-Seminar will be the Mannheim SuParCup'99 Awarding session. The award will be performed by Gerhard Widder, Lord Mayor
of Mannheim and by Günther H. Oettinger, Chairman of the CDU-fraction in the state parliament of Baden-Wuerttemberg.
|
| In 2010 nearly every organization will have its own petaflop supercomputer
|
|
|
"In 2010 nearly every organization will have its own petaflop supercomputer ". This bold statement was made by Gordon Bell, the first-day key-note speaker at the Mannheim Supercomputer Conference.
Gordon Bell, senior researcher in Microsoft's Telepresence Research Group, which is part of the Bay Area Research
Center (BARC), forecasted a future, building upon clusters.
Companies that currently have 10.000 processors in all the workstations and servers they own, will have 100.000 in 2010; and they will be faster. Clustering these machines will provide tremendous
supercomputing power.
However, the programming
remains the major barrier for most applications.
|
|
|
| The"Quo vadis?"
of high-end computing
|
|
|
Greg Astfalk is the Chief Scientist for Hewlett-Packard's
High Performance Systems Division (HPSD) where all thecompany's
mid-range and high-end Unix servers are designed and built.
In the afternoon
session "Meet the Chief Technology
Officers" at the Mannheim Conference, Mr. Astfalk shared his impressions on 2010 supercomputing with the audience.
A big warning in advance: any resemblance in this speaker's talk to real HP plans is purely coincidental.
The conference attendants were invited to listen to some general comments on the hopefully prospering future of technical
computing.
|
| How the "coupling problem"
has to be overcome
|
|
|
As the responsible person for the assessment of Sun's technological
investments and as director for the activities of Sun Laboratories and associated development
programmes, Dr. Greg Papadopoulos was invited at the Mannheim
Supercomputing Conference to offer his views on High Performance Computing.
The sticky edges of parallelism reside in the diffulty to break
the computing process for a specific problem into parallel chunks.
How to try and lay-out the data
for a problem on memory/disk and especially how to cope with these issues simultaneously?
|
| A supercomputer as reliable and general purpose as a T-Ford
|
|
|
Dr. Burton Smith, Cairman and Chief Scientist of Tera Computer Company tried on the shoes of Henry Ford during his talk at the Mannheim Supercomputing Conference.
The big problem today in modern processing can be described as latency disease.
It is quite easy to localize the symptoms in the existence of too much registers, massive caches, multi-way issue, and only one sigle programme counter.
Particularly in parallel machines, the problem is very urgent.
The excuse for putting off possible solutions relates to the affordability cry.
Experts claim that there is no market for processors which are more suited to high performance computing.
Negativism never has been able to create a market.
Here is where Henry Ford comes into the story...
|
