The race for computer power and speed has been underway since World War II. It accelerated tremendously in 1971, with Intel’s introduction of the first microprocessor. Soon afterward, one of Intel’s founders, Gordon Moore, observed that price-performance was doubling every 18 months, and that trend became known as Moore’s Law.
Since then, Intel has held a solid lead in this industry. It dominates the processor market, and has increased speeds year after year until reaching today’s commercial benchmark of 3.8 gigahertz for its current Pentium 4 processor.
But all of this has been achieved using virtually the same architecture as the original microprocessor, the so-called X-86 architecture pioneered by Intel. The improvements in speed and power have come simply by refining the same technology. This is the same approach to incremental innovation that car manufacturers have used to refine the internal combustion engine over a century to make faster, more efficient cars without ever jumping to a dramatically different design.
Now, according to the Wall Street Journal, a radical innovation, the multi-core processor, is causing an upheaval in the industry. The new technology was unveiled by both AMD and Intel in San Francisco at the Developer’s Forum last September. It promises to turn the chip industry, and perhaps Moore’s Law, on its head. A little background is in order.
Some time ago, engineers at places like IBM, Hewlett-Packard, and Sun Microsystems began to realize that they could make more powerful and efficient servers if they used several processors instead of just one. Software vendors began writing programs for those multi-processor servers, laying the groundwork for future developments. As traditional X-86 chips began to hit their maximum speeds, and problems with heat build-up began to plague the faster ones, engineers peered into this promising area for the consumer electronics markets.
They discovered a better method than using multiple chips, which quickly get bulky and expensive. The new idea was to put more than one processor core on a single chip. This has tremendous cost advantages. It allows the cores to share common components, such as cache, and it speeds up inter-processor communications because all the components are on the same chip. In addition, the multi-core processors produce far less heat, and use far less power, than multiple processors.
Single processor chips can only do one thing at a time. For example, if you’re trying to process video, it takes up most of your processor’s power, and the whole system slows down to the point where you can’t do anything else. While the cores in a multi-core processor may not go any faster than a traditional chip, they can all do different jobs, so the apparent speed of the system is much, much faster ? in some cases, as fast as a supercomputer. Whatever the actual speed, Business 2.0 called multi-core processors one of the hottest new technologies of 2004.
In practical, day-to-day terms, this means that new software, originally developed for high-end, multiple-processor servers, can be tweaked to address the multiple cores in a single new-generation processor. If you’re using a game program, for example, the software would direct the video portion to one processor, the audio to another, and so on, providing a much faster and more realistic experience. And this is where the big opportunities lie: Everyone in the multi-core game is jockeying for position in the enormous home entertainment market.
In the wake of the AMD and Intel announcements, IBM, Sony, and Toshiba unveiled details of their joint multi-processor chip effort, code-named the Cell. The companies claim that it can rival supercomputers in powering everything from video games to business computers.
According to the Associated Press, Sony will introduce its new PlayStation 3 featuring a Cell chip, while Toshiba will use it for high-end television, and IBM will start marketing Cell workstations ? all within the 2006 calendar year.
The Cell is based on IBM’s PowerPC architecture, which is used in Apple computers today. While Intel’s traditional chips can carry out two instructions, the Cell can carry out 10 and run multiple operating systems.
According to the Pittsburgh Post-Gazette, IBM, Sony, and Toshiba’s announcement of the Cell processor has in effect fired a shot across the bow of Intel and AMD in the coming war for the universal home entertainment market. The Cell, for example, can perform the roles of a video game console, a television, and a general-purpose computer all at the same time.
To compete directly with IBM’s cell processor announcement, Intel revealed that it had completed its first product runs of dual-core processors based on Pentium chips. But the Cell has 234 million transistors compared with Intel’s 125 million. It’s also said to beat Intel’s clock speed by 0.2 gigahertz.
Since the 1970s, one company after another has attempted to unseat Intel as the leader in this field, but so far no one has succeeded. For example, IBM teamed with Motorola and Apple to push the PowerPC, which is still in use in Apple Macintosh computers and IBM workstations, but it failed to change the paradigm.
In 2000, Transmeta launched its Crusoe chip to challenge Intel in the notebook market, but gained only marginal acceptance before fading away. Intel’s Centrino, launched in response, is now dominant in mobile technology.
Another high hurdle for those taking on the chip giant is that chips have always needed backwards and forward compatibility. A truly new architecture will not be compatible with old systems and software, so whoever wants to be the challenger will have to set the performance bar high. That may be precisely what happens when users see photo-realistic graphics powered by a Cell chip in the PlayStation 3.
In the meantime, both AMD and Intel will start introducing dual-core processors for workstations as soon as possible, with true multi-core designs to follow. Intel has already announced that its future lies with multi-core processors, so it’s pretty clear that the Pentium legacy will soon be left behind in favor of a new architecture. The question is, whose architecture will it be?
Based on this compelling trend, here are six forecasts for your consideration:
First, in the short term of the next five years, dual-core and multi-core processors will be an extremely hot topic in technical circles as the various competitors race to establish a clear technical lead in performance, price, and features. This will go on largely outside the public forum. In the meantime, those increasingly sophisticated processors will battle it out in dedicated applications, such as CAD workstations and gaming devices, which do not need to be fully backwards-compatible.
Second, in the five- to 10-year time frame, a clear winner will emerge, with the most likely scenario being the Intel-AMD team going up against the IBM-Sony-Toshiba alliance for a fight to the death over dominance of the next-generation paradigm for computer chips. It will be a winner-take all battle, as it was with Intel’s long dominance of the single-core processor market.
Third, at that point, a secondary battle will emerge for dominance of the home entertainment market with products that integrate all functions into one unit, from gaming to television to Internet to home computing. That will fragment the makers of consumer electronics into numerous product and marketing schemes, with a few big winners and many intermediate players finding niche markets. The dominance of Sony and Toshiba in games and television, respectively, is not a foregone conclusion by any means.
Fourth, the nature of software licensing will change, with Microsoft positioning itself for dominance, at least in the short term. Traditionally, server licenses are issued on a per-processor basis. Many software vendors see multi-core processors as an opportunity to shift licensing to a per-core basis, upping revenues. Microsoft has announced that it is sticking to its per-processor licensing scheme, thus squeezing the competition. In other words, if you use Microsoft, you pay the same price, no matter whether you use a traditional single-core processor or a new one with five, 10, or 20 cores. Sun Microsystems is experimenting with a per-person license. Whoever wins, licensing will be a big issue as multi-core processors emerge into wide use in the market.
Fifth, in the 10- to 20-year time frame, expect the multi-core processors to compete head-to-head with programmable logic devices ? or perhaps mutate into a new form of programmable logic device. As you will recall from the March 2005 issue of Trends, programmable logic devices, or PLDs, are chips that have no predefined architecture but can be re-programmed on the fly for many functions. Some multi-core chipmakers will undoubtedly begin putting programmable modules in some ? if not all ? of their cores because of the extreme flexibility this affords. Again, a winner will eventually emerge as the dominant maker of a new standardized chip architecture that will supplant the Intel X-86 architecture in a wide variety of consumer and professional products.
Sixth, in the 20- to 50-year time frame, even this forthcoming generation of chips will be supplanted. By then, nanotechnology, biochips, and quantum computing will render traditional silicon technology obsolete.
References List : 1. The Wall Street Journal, May 13, 2004, “Intel Making Big Design Strategy Shift with Multicore,” by Don Clark. ⓒ Copyright 2005 by Dow Jones & Company. All rights reserved.2. Business 2.0, October 2004, “The Hottest Technologies of 2004,” by Michael V. Copeland. ⓒ Copyright 2004 by Business 2.0 Media, Inc. All rights reserved.3. Associated Press, February 7, 2005, “IBM, Sony, Toshiba Unveil Microprocessor Challenge to Intel,” by Matthew Fordahl. ⓒ Copyright 2005 by The Associated Press. All rights reserved.4. Pittsburgh Post-Gazette, February 8, 2005, Tech Giants See Answer to Intel Engineers in New Microprocessor,” by Matthew Fordahl. ⓒ Copyright 2005 by PG Publishing. All rights reserved.
