New and innovative computer designs -- which many experts predict may spark a new revolution in data processing -- will be slow to be developed and have only a negligible market impact over the next decade, according to a top International Business Machines Corp. executive.
"At risk of sounding like the conservative IBM type," said Herbert Schorr, group director of technology and products at IBM's Information Systems Group, "the prognosis for a new architecture coming in to replace existing computer architectures is rather dim.
" . . . The bottom line is there's not going to be much change over the next 10 years."
Speaking at the "Future Generation Computers" panel here at the National Computer Conference, Schorr dismissed the widely held belief that new computer designs are needed to cope with sophisticated artficial intelligence programs now being developed.
"The real problem facing us is how to integrate that software in our existing operating systems," he said.
With roughly $50 billion a year in revenue, IBM is the world's largest and most influential computer company. IBM's large mainframe computers and personal computers have become industry standard for their markets. Schorr's comments reflect IBM's coventional wisdom as to how computer technology for business is expected to evolve.
However, a growing body of computer academics and entrepreneurs insists that new forms of computer designs and architectures will swiftly and dramatically improve computer/cost performance -- much as minicomputers and personal computers revolutionized the industry.
They point to the defense department's Strategic Computing Program and the Microelectronics and Computer Technology Research Consortium as efforts designed to develop and promote new computer architecture.
Indeed, several young companies at the NCC -- including Encore Computer Corp. and Sequent Computer Systems Inc. -- are now selling what are described as the forerunners of this future computer architecture.
The key to this architecture is the use of multiple processors running simultaneously in a computer to solve problems. Traditionally, computers have relied on a single high-speed processor to solve a problem. The faster the processor could process information, the more quickly the problem could be solved.
The logic behind multiple-processor computers is that two processors could solve a problem twice as fast and three processors could process the problem three times as fast, if the problem could be broken into the appropriate number of pieces.
However, designing a computer that lets the multiple processors effectively communicate with each other as they solve a problem has proven extremely difficult. In addition, writing software for multiple processors to solve problems "in parallel" has proven particularly challenging.
"When it comes to general purpose parallel processing computers," said IBM's Schorr, "the results appear to be extremely unfavorable or negative."
Yet companies such as Sequent and Encore report that their multiprocessor machines can now solve certain kinds of problems up to 50 times faster than traditional machines and at a lower cost.