New supercomputers


ISSN: 0368-492X

Article publication date: 1 April 2000



Rudall, B.H. (2000), "New supercomputers", Kybernetes, Vol. 29 No. 3.



Emerald Group Publishing Limited

Copyright © 2000, MCB UP Limited

New supercomputers

New supercomputers

Keywords: Automation, Cybernetics, Research, Technological developments

The need for the supercomputer

There is so much concern these days about the future of the Internet and the development of PCs that many so-called computer experts have shown little interest in the development of the supercomputer.

The time, we are told, of the super-cruncher is over, it is now the "cyber-age" when only the PC and its communications links are of importance. Most cyberneticians and systemists know better and the need for super-fast computers that are capable of tackling the enormous problems that we have yet to solve is as great as ever.

The reason is quite simply that most of today's computer users do not have problems to solve that require tremendous computer power. A recent survey showed that a high percentage of computer "experts" of today's generation of graduates had never tackled a "real" problem on a computer that needed the use of a sophisticated programming language. Someone else had always provided the software that was required and invariably the machine used was a standard PC. They have yet to appreciate that some scientists have problems that existing software cannot define and current machines are not powerful enough to handle. Such problems, for example, occur constantly in research modelling. Researchers in biotechnology require to produce research models involving human proteins and the availability of faster supercomputers could speed up the design of new drugs for the pharmaceutical industry. Many other fields of endeavour are struggling with problems that cannot be tackled unless there is a breakthrough in the design of our supercomputers so that their performance is vastly improved.

New breed of supercomputer

There are now indications that the main manufacturers of large main frame machines are developing computers with such greatly enhanced performances. IBM has announced that is investing some $100 million (about £60 million), in its development project to produce a new supercomputer. It is to be, the company claims, 500 times more powerful than today's fastest machines. It is called Blue Gene and its tremendous computing power is expected to be 1,000 times as great as IBM's Deep Blue. It will be recalled that IBM's Deep Blue was the machine that taxed the world's chess masters and beat the world's champion, Garry Kasparov, in 1997.

It is also of significance for today's "expert" PC users to note that the planned machine, Blue Gene, is to have a power which is two million times as great as a current desktop PC.

IBM vice-president, Ambuj Goyal, said of the machine: "we think a tremendous gain in performance will be made possible by the first major revolution in how computers are built since the mid-1980s".

The company has described the new supermachine as having more than one million separate processors - each capable of one billion operations per second. The whole set of computer processors will work together in a single 2,000 sq. ft machine.

At a New York launch IBM's senior vice president of research, Paul Horn, displayed a model of the circuit boards that will go into Blue Gene. Each will contain 64 1 gigaflop chips, and about 1,000 boards will be built into the machine. IBM said that it expects to take between four to five years to build the new supercomputer which, the company claim, will perform one million billion mathematical operations a second. One innovative feature of this powerful machine is that it is to be "self-healing", that is, it is designed to be able to detect failing components, seal them off, and direct the work elsewhere.

But will it be powerful enough to tackle some of the enormous calculations that await processing if we are to deal with even the tip of the iceberg that has emerged in our research? Simulating the building of proteins in the body, for example. One report from an

observer at the launching of Blue Gene assesses its challenge as "daunting". Proteins, it was said, start out as long strings and fold themselves to a specific shape that determines their function in the body. This process takes a fraction of a second. It was calculated that even with the great power envisaged for IBMs Blue Gene it would take a year to calculate how such a typical protein folds itself.

There are, however, many reports of projects that are designed to produce new "revolutionary" computing machines with new technology and new architectures which may well change the way in which we tackle these enormous calculations. These systems are still speculative in concept and design. IBM has, however, shown us the blueprint of an actual machine that is to be built. There is no doubt that it will make an enormous contribution to science and in particular to our understanding of so many aspects of science that will illuminate the "secrets of life".