A new mini stirs the industry
DEC's powerful system moves minicomputers closer to IBM's markets
November 7, 1977
This week Digital Equipment Corp. announced a new computer at its annual meeting in Boston, and President Kenneth H. Olsen, normally a shy, conservative executive, threw caution to the winds. He hailed his new VAX minicomputer as "probably the most significant interactive computer of the last decade." He also called it "a milestone equal to the original PDP-11." That is something like an oil company president describing a new oil field as the richest find since Saudi Arabia. More than 50,000 PDP-11s have been shipped by DEC in the past five years, fueling the Maynard (Mass.) company's 400% increase in sales since 1973 to $1.1 billion in the year ended June 30.
There seems to be no doubt that Olsen has an important new product. Reports speculating on the DEC minicomputer have circulated throughout the industry for more than a year. But even when it was announced, there was a great deal of disagreement over what kind of impact VAX would have on DEC and the computer business in general.
Most important, the VAX 11/780 -- "obviously named by a committee," grumbles one competitor -- operates with a 32-bit word, the same size that the big IBM mainframes use. That means VAX can handle 32 bits of data at once, making it potentially far more powerful and faster than any 16-bit minicomputer, which is currently the industry standard. But the price tag is stiff: $130,000 for a barebones model and up to $1 million for a large system.
The fact that a minicomputer has a 32-bit word will mean very little to most people, and even those who operate with sophisticated programming languages that now come with most minicomputer systems. But the longer word is important, because it makes for vastly more powerful minicomputers. VAX, for example, can address directly up to 2 million bytes (a byte is 8 bits, or one character) of computer memory. The standard 16-bit mini, or the other hand, must use gimmicks or special hardware to go beyond a relatively paltry 128,000 bytes. And memory size is becoming important in a growing number of minicomputer applications.
DEC contends, and many in the industry agree, that a lot of minicomputer users are approaching performance limitations on 16-bit machines. These users want to stay with the less costly, interactive minus rather than move up to the big mainframe computers, which typically are too big for what they want, both in capacity and in price. "The 32-bit machine is a natural progression for DEC," says David Methvin, president of Computer Automation Inc., an Irvine (Calif.) mini maker. "And it's probably a good thing for the industry," he adds, "since it gives minicomputer uses one more option before they have to leap on up to mainframe."
So the 32-bit minicomputer is the industry's effort to combine what DEC calls the power and performance of mainframces with the flexibility and low cost of minicomputers. As such, the new machine will provide much of the push in the industry's upward expansion into new markets.
DEC, along with just about everyone else, expects a strong market for 16-bit computers to continue indefinitely. "If this were going to impact the 16-bit market, I seriously doubt that DEC would do it," Computer Automation's Methvin says.
But the 32-bit machine could ultimately dominate the market. Like it or not, the 32-bit race is on for the minicomputer industry, declares Samuel H. Bosch, product marketing director for Systems Engineering Laboratories Inc. (SEL). "DEC and Data General set the standard in the 16-bit machines," adds an executive at Datapoint Corp., "and they have a good shot at setting a new standard in the minicomputer industry at 32 bits."
DEC will not be the first company to market a 32-bit minicomputer. Two small competitors, Perkin-Elmer Corp.'s Interdata Inc., and SEL, have been selling the big minis for several years. Both companies expect to introduce second-generation models within the next year. And Data General Corp., DEC's closest and most bitter rival, also is hard at work on a 32-bit system.
As the major producers begin shipping the larger minis in the next year or two, the market is expected to grow sharply. William A. Feldman, marketing director at Interdata, bullishly predicts that 32-bit systems will grab $2.5 billion of a total mini market of between $5 billion and $6 billion in 1980. That estimate, however, is well above those of most observers who expect the growth to be more gradual.
Many in the industry figure that they know the real reason why DEC is going to a 32-bit word system. "DEC has the semiconductor companies coming up from the bottom [with the computer-on-chip] and IBM coming down from the top [with its new minicomputers]," says James G. Treybig, president of Tandem Computers Inc., a Cupertino (Calif.) minicomputer maker who is not interested in copying DEC's move. "The only reaction DEC has is to build a more powerful system," he says, "and they can't build a more powerful 16-bit system."
But DEC's Olsen says that VAX is strictly a minicomputer, and DEC has no intention of competing head-on with IBM. "The world is waiting for us to say, 'we're going after the IBM mainframe business', but it just isn't so," he says. "The idea of entering IBM's markets never entered our heads as we designed the VAX," he insists.
Whether it did or not, the collision with IBM in some mainframe markets seems inevitable as the mainframe and minicomputer worlds move closer together. The VAX will replace mainframes in some applications, or take away some jobs from mainframes, acknowledges Julius L. Marcus, vice-president of DEC's Information Systems Group.
There are still many other new applications out there for the fast-growing minicomputer business. For VAX, they will include such jobs as larger scientific systems, data communications processors, and network processors in distributed-processing systems. And there will also be the applications that no one has even thought of as yet. "I'll lay odds the mini makers continue to find things to do that IBM doesn't," declares William H. Davidow, general manager of Intel Corp.'s Microcomputer Div., which dominates the computer-on-a-chip business.
Davidow is just one of many semiconductor industry executives who have their eyes on the low-price end of the minicomputer market. The industry's microcomputers are now principally 8-bit systems that are slow and severely limited in power. But not for long. "Next year there will be at least one 16-bit system out that will be 10 times faster than today's 8-bit systems," predicts Davidow, who most likely is talking about a new Intel product.
While this kind of low-cost competition from below may spur minicomputer companies to move to higher-performance ground, Davidow figures that they will do so anyway. The reason is that it probably costs only about 15% more now to make a 32-bit system that a 16-bit machine because of the steady decline in hardware costs.
But the success of DEC's new super mini will be determined not by its hardware but by the software that the company will provide for it. And developing that not only takes years but large sums of money. "You're talking entirely new software, so the 32-bit machine is a damned expensive proposition," says Neal F. Young, market planning director at General Automation Inc., an Anaheim (Calif.) minicomputer maker.
Interdata is a good case in point. In 1973, with great fanfare, the Oceanport (N.J.) company introduced a 32-bit minicomputer, and the product promptly went nowhere. The problem was a lack of sufficient software, and it took several years and a heavy investment by Interdata to correct it.
The advantages of a 32-bit system disappear if the software is not developed to utilize fully that word size and system architecture. So DEC competitors, as well as industry analysts, see VAX primarily as a compatible extension of the PDP-11 family. Initially, at least, the new DEC system will undoubtedly go primarily to customers that have out-grown their PDP-11 systems.
The Signal Processing Facility at TRW Inc.'s Defense Systems Group in El Segundo, Calif., for example, is running PDP-11s and wants to buy a VAX system. The facility does both image and signal processing and has to manipulate a lot of data very rapidly. "VAX looks like a painless avenue of growth," says Roger Vossler, facility manager, because it is compatible with his PDP-11 software. For less than $200,000 the VAX system offers "a lot of performance for the dollar," he says.
The initial software that DEC will offer for VAX, such as FORTRAN engineering language, makes it clear that the company will move slowly with its 32-bit system, moving first into higher performance areas of its traditional scientific, industrial control, and engineering "number crunching" markets. One early buyer of $200,000 VAX system was Pittsburgh's Carnegie-Mellon University. "VAX will have quite an impact for scientific users," says John A. Pople, professor of chemistry. The new machine came along just at the right time for the chemistry department, since it needed its own dedicated computer. It has been struggling along by sharing the school's Univac 1108 mainframe system. "Primarily we need to address directly a lot of memory and to jump about between arrays," Pople says, "and the 16-bit computer really is quite hopeless for our needs."
But such scientific business is paltry compared to the massive and fast-growing business market for minis. "If DEC had come out with high-level COBOL, software, they could have hit IBM quite heavily," says the Datapoint executive. DEC is hard at work developing the 32-bit COBOL language, but the job is a major undertaking, and for now, commercial users will have to get along with the PDP-11's version of COBOL, which negates most of the hardware power of VAX.
"We can't do everything at once," Olsen says. The DEC president is convinced that the three years that his company spent in designing VAX will pay off in longevity of the basic architecture, providing a solid base for many new systems as the years roll by. "The basic architecture is critical to our future plans," he adds.
Not everyone is convinced that the 32-bit architecture is the only way to expand the minicomputer industry's marketplace upwards. Two companies already have moved in the VAX direction with models that look like 32-bit systems to users but have 16-bit architecture. Last week Wang Laboratories Inc. introduced its 32-bit "lookalike," and, even more than DEC, tried to make the entire architecture behave as much like an IBM mainframe as possible. "The user doesn't care as long as it works," maintains Peter McElroy, product manager. The idea, he says, is to extend Wang's low-cost business systems into the $50,000 to $150,000 range by making them "comfortable" for IBM users to accept.
Another 32-bit lookalike comes from Prime Computer Inc., of Wellesley Mass., which is growing at better than 100% annually because of the success of that product. At one Prime installation, a group of utilities is doing load-flow analysis at 10% the cost of doing it on an IBM 370/158, the predecessor system, claims David C. Hetrick, planning director for Prime.
New techniques that take advantage of new semiconductor developments can do the job as well as doubling the word size, says Marco Negrete, engineering manager for Hewlett-Packard Co.'s Computer Systems Group. Herbert J. Richman, marketing vice-president at Data General, says that "a 32-bit product is interesting, but it's not really the wave of the future." There are very few situations that cannot use a 16-bit system, he claims. "No 16-bit software will work on a 32-bit system," he adds, "and until Data General gets a system we're satisfied with, both hardware and software, we won't announce it."
If any case, it seems clear that IBM will continue to have a lot of scrappy little -- and not so little -- companies as competitors, as rapid changes in technology tend to obliterate old boundaries in the computer business. Tandem Computers, for example, is linking 16-bit minicomputers to build multiple-processor systems that are capable of outperforming the biggest IBM mainframes. "You don't need the 32-bit word length to beat IBM," claims Tandem's Treybig. "We're taking major accounts away from IBM -- and DEC, too."
GRAPHIC: Picture, DEC's Olsen and his VAX minicomputer: Will the powerful system set a new industry standard? Ivan Massar/Black Star
Copyright 1977 McGraw-Hill, Inc.