| Previous | Table of Contents | Next |
Speculating about what the computer industry will be like in three to five years is a bit like guessing who will win the World Cup, the Formula 1 Championship, or the Super Bowl in the year 2001. It is not impossible to guess these things, which probably explains why so many people have an opinion and are usually willing to share that opinion. But aside from accumulating a bunch of opinions, to speculate about such matters may or may not be of much of value. Computer consultants regularly predict five years into the future, even though their track records for such predictions are often quite poor. Still, we do know a great deal about the technologies that will be available in the future, so I want to offer my opinion about the computer industry after the year 2001.
You may wonder why I select the year 2001, rather than the year 2000, as my starting point. Aside from my being a purist in pointing out that the 21st century doesn’t actually begin until the year 2001, there are other reasons for looking beyond 2000. Beginning in 1998 and continuing through 1999, many companies will slow their adoption of new technologies, especially if the adoption takes a great deal of effort, because of the most-dreaded date in the computer industry — January 1, 2000. It is difficult to predict the magnitude of the year-2000 problem, so most businesses are unwilling to make major changes to their computing environments in the interim unless those changes are absolutely necessary. And even companies that are taking on massive changes are finding it difficult to hire people with the right skills, because many of those people are busy solving year-2000 problems.
As we all know, when the calendar year changes from 1999 to 2000, many application programs and some operating-system programs will fail to run correctly. These programs will think the year is 1900, because they use a two-digit encoding for the year. All the incorrect results these programs produce are likely to create a great deal of turmoil around the world.
Many companies have already put enormous efforts into solving their year-2000 problems. Software and operating-system vendors, too, have been making their products year-2000 enabled. For example, all AS/400 system software and many of IBM’s business partners’ software has already been certified as ready for the year 2000. By the year 2001, we should all be back to normal, looking to see how new technologies can benefit our businesses.
I start my speculation with a topic about which it is easy to make predictions: processor hardware technologies. The laws of physics govern hardware, so it is possible to predict with some certainty how a given hardware technology will evolve over time. What I cannot predict with very good accuracy is which technologies will find their way into our computer systems and how the technologies might be used.
In Chapter 2, I introduced Moore’s law (which states that the number of transistors it’s possible to put onto a single chip will double roughly every 18 months). Many of us in the industry believe that Moore’s law will hold for at least another 15 years, and that means the computer industry is not likely to abandon the CMOS technology for a long time. It also means that the number of circuits on a silicon chip will continue to double every 18 months as a result of our ability to shrink the size of the transistors on the chip and, therefore, to pack more of them onto a single chip. The performance of these chips will increase as the distance between the transistors becomes smaller and smaller. Electrical signals on a chip travel at nearly the speed of light, so if the distance the signals have to travel is reduced, the performance will increase.
We often take for granted the advances that have been made in hardware technologies. The past 30 years have brought tremendous improvements in performance, cost, and power consumption for computer hardware. To put this into perspective, suppose the automobile industry had made the same progress in the past 30 years. If this were so, you could buy a new Porsche for about $2; the car would go faster than the speed of sound and could travel more than 1,000 miles on less than an ounce of gasoline — make mine a red one!
In this section, I speculate a little about the future of AS/400 hardware and what we might expect in the next 5 to 10 years. As with any such speculation, there is no guarantee that anything I predict will come true, but it is fun to do. So fasten your seat belt; our new Porsche is rapidly approaching the speed of sound as we drive toward this possible future.
The AS/400e series will use PowerPC processors for the next several years. As we saw in Chapter 2, the third and fourth generation of PowerPC processors developed in Rochester will carry the AS/400e series through the Version 4 timeframe and beyond. Models of the RS/6000 are using these processors. (IBM discussed this idea of processor convergence between the AS/400 and RS/6000 for commercial processing applications from the beginning of the PowerPC work but could not achieve that convergence in time for the first generation of AS/400 processors.)
The first- and second-generation AS/400 processors supported only the tags-active mode of operation. All third- and fourth-generation processors implement both the tags-active and the tags-inactive mode. As we have seen, these processors also support industry-standard I/O interfaces and can be used to run OS/400 or any other PowerPC operating system.
In Chapter 2, I discussed the original design for a high-end converged processor called Belatrix, named after a star in the constellation Orion. To review, that single design, which was originally planned to satisfy both NIC and commercial requirements, was overly ambitious for the schedule and was cancelled. As a result, the laboratory in Austin began to design its version of Belatrix for NIC applications, and Rochester began work on its version, called Northstar, as the first in the family of fourth-generation processors. This family of processors is optimized for commercial-processing workloads running on either the AS/400e series or the RS/6000.
| Previous | Table of Contents | Next |