IBM presentations are so stylised and IBMers are so cryptic that what is said – and what is left unsaid – has to be dissected with the same diligence as was once devoted to entrails, and more recently to photographs of Soviet leaders lined up for the May Day parade if one is to have any hope of divining the true situation – or indeed what the company is really saying. A classic example is last week’s New York presentation of the state of play with the RS/6000 (CI No 1,469), and taking everything that was said together, the conclusion has to be that IBM’s second attempt at a convincing Unix machine has got off to a very good start as far as the IBM AS/400 and 9370 camps are concerned, an extremely disappointing one from the point of view of the true Unix believers within IBM.
When an IBMer says that IBM will ship substantially more than 15,000 RS/6000s by the end of the year, it doesn’t mean that IBM will in fact ship 20,000 to 25,000 by the end of the year, as some gung-ho commentators talking their book want us to believe. It can mean either that the speaker is confident of getting 17,600 away by the end of the year if things are going reasonably well, or that he’ll bet his life on it being 15,560 so that in round figures it’s 16,000. That gives us a pretty fair idea of what IBM expects to do. Then IBM says that initial orders came in at the high end of expectations. That’s bad. For public consumption, IBM always sets its expectations for new products ludicrously low so that it can say that demand has greatly exceeded our most optimistic expectations. This one hasn’t even met what will have been a very modest internal expectation. But the killer comes with Bill Filip saying firmly that everyone that wants one will get one in 1990. And this is for a machine that was intended to have been announced last October, was announced in February and has had a mass of marketing hype behind it ever since. By contrast the AS/400 wasn’t announced until June 24, yet IBM managed to get 30,000 of the things away by the end of that year. It is of course grossly unfair to compare the RS/6000 machine with the AS/400, because there was very substantial pent-up demand for that box from System 38 users desperate for more power. And yet IBM has had six months already to build demand for what was perceived, when it was announced, as very good value for money in raw processing terms, and is still very well regarded conceptually. Yet from what IBM was saying last week, the customers are not exactly battering down the doors to get what was intended to be IBM’s hottest low-end and mid-range property this year. What demand there is appears to be coming primarily from the design automation market – but unfortunately, IBM is not able to ship the models most suited to such applications – only the low-end 320, 520 and 530 are shipping now, the high-end 540 and 930 servers will follow shortly, but the model with the hot graphics, the 730, won’t be shipping until November.
And already, IBM is beginning to fall behind: the companies that it has to beat, Sun Microsystems, Hewlett-Packard Co and a host of smaller vendors, have added machines that eat hungrily into the price-performance advantage IBM claimed for the RS/6000 in February. IBM is promising to double performance every 12 months, but the first iteration of that process has to happen this month or IBM will be falling behind the competition and its ability to match rhetoric with performance will come into question again. Such promises from IBM are anyway of little value – when it launched the AS/400, it promised to double performance every two years, and to keep that promise it needs to ship an AS/400 50% faster than the Model 70 next month. Any sign of it? But that is just the hardware story: things look much worse when software comes under the microscope. IBM has committed to moving from AIX 3.1 to OSF/1 on the RS/6000 at some time in the future, but it is unable to say when, which renews suspicions that the Open Software Foundation will be unable to ship anything li
ke a solid operating system in November – if it were only a question of waiting until November, it’s difficult to believe that Siemens AG would now be planning to go to Unix System V.4 on its forthcoming Intel iAPX-86 family machines. Nor can IBM mix it in the gutter with the really competitive diskless workstations from the likes of Sun: You need to be able to path in orthe New York meet last week. These functions are not in AIX 3. Supposed to be there, we would have liked to have them there… They’ll be coming. And on low-cost machines in general, This is probably stuff we’re going to be delivering in the first part of 1991. If the guys in the lab can get the software right. But unless and until IBM drops its commitment to OSF/1, all this is academic anyway: what serious software vendor or large end user is going to devote substantial resources to developing major applications for AIX 3.1 when the threat that the applications will have to be converted to run under OSF/1 is looming. Far better wait until OSF/1 is here before making a commitment. All of that implies a two-year hiatus while the developers that matter sit back and wait to see what happens. And all IBM’s recent history insists that that is the one thing that IBM cannot allow to happen. If the RS/6000 takes two years to get off the ground, it will never happen.
Because IBM has never succeeded in rescuing a new hardware product that was perceived as a failure after its first 18 months: the list of failures is an embarrassingly long one Series/1, the 5100, the 8100, System 23, the RT, the 9370, the Portable Personal Computer and all subsequent portables, even the excellent System 38, which needed to be transformed into the AS/400 to assume the aura of success. And the winners over the same period? System 34, which turned into System 36 for no good reason, the 4300, and the Personal Computer and desktop successors. And when it comes to entry into a market completely new to IBM, only the Personal Computer rates as a success – much too big a success for many IBMers. And how did the Personal Computer differ from all the other machines? The operating system was bought in and owed nothing to IBM’s vast but fatally flawed software development effort. There is still time to save the RS/6000, but less than IBM and fans of its Unix effort would like to believe. And the signs now are that triumphant success demands nothing less than that IBM swallow its pride, dump AIX, forget OSF/1, and buy a licence to Unix System V.4. Otherwise, the AS/400 camp at IBM will be able to sleep easy in its bed for three or four years yet. But then isn’t that the scenario, in its heart of hearts, that Armonk really wants to see anyway?