The 80386 chip promises to become the biggest moneyspinner of all time among microprocessors, and fortunes will be made and lost on the it. A host of companies are crucially affected by availability of the chip, and boardrooms on either side of the Atlantic are populated by executives with churning stomachs and bitten fingernails. The Microbytes Daily newswire has been assessing the combination, and finds little immediate cheer for the host of manufacturers anxiously awaiting 80386s in the number they need – and finds that Intel Corp is playing a decidedly dangerous game in its efforts to keep the entire business for itself. Several weeks ago, Intel announced that, because of a 32-bit arithmetic execution problem, the 80386 would have to be remasked and production would slow down, putting chips in short supply until the end of the year. By that time, it said, the company would have delivered between 500,000 and 1m. But while Intel is still saying that it remains on track to hit this goal, it reports that within the past few weeks, demand for the 386 has tripled and that greater shortages and longer delays than expected would be likely, adding that full production of the remasked chips is not expected before the end of July at the earliest.
In worse shape
But at least one industry watcher contacted by Microbytes reckons that by the end of the year, the supply and demand situation will be in worse shape than Intel says. At this time, the only way they can meet the commitments they’ve made is to dump all the parts on customers all at once in the fourth quarter, and companies won’t have any chance to gear up manufacturing in a normal fashion. If the faulty arithmetic-logic circuits are avoided, the 80386 effectively becomes a 16-bit chip working in the real only and not the protected mode. Customers currently report that they can get all of the 16-bit 386s they want but the remasked ones are under allocation. On the other hand others believe that Intel’s main customers, IBM, Compaq, PCs Limited, for example – are getting all the 80386 chips they want, while smaller customers, particularly those in the Far East, aren’t getting any at all. Others affected by the low availability of the 80386 include other chipmakers like Chips & Technologies, the company that puts together those super-attractive PC-DOS cloning kits. We can’t ship 386 chip sets without 386s out there, it says. The market is being held back just at the time it should be ramping up. What of Intel’s efforts to keep the 80386 all to itself, and not allowing even its contracted second source, Advanced Micro Devices, have a crack at the part? An Intel official, Gene Meieran, is quite open about the policy, saying that single- sourcing will be increasingly important in the semiconductor industry, and that customers prefer a small number of reliable suppliers instead of a large number of questionable suppliers. Intel elaborates, saying Instead of going to second sources, we are spreading our fabrication plants out around the world. Intel currently has two 80386 fabrication plants in operation, one in Livermore, California, and another in Jerusalem, Israel, and it plans to open one in Oregon by the end of the year and a fifth in Albuquerque early in 1988.
Dispute with AMD
Meantime the dispute between Intel and Advanced Micro over AMD’s claim that it has the right to second-source the 80386 appears to be no closer to resolution than before, and neither company wants to forecast when it will be resolved. The dispute began in April when Intel unilaterally cancelled the 10-year technology exchange agreement with AMD initiated in 1981, alleging non-payment of some royalties. AMD wasn’t buying that one, and the issue is now in binding arbitration, as called for in the original agreement. AMD’s declares that it provided devices that fit Intel’s needs and thereby built up points – the basis on which a second source gets to move on to the next key part. It is demanding masks to both the 80386 and to the 80X87 maths co-processors. But AMD acknowledges that In
tel has enough capacity to make the 80386 itself – and that’s what it’s going to do. But apart from the possibility that AMD will come out the victor in the arbitration, and the problems caused by the bug and the need to remask, Intel has two other nagging concerns to worry about. The first is an agreement it had little choice but to sign with IBM. Some Intel customers reckon that one of the real reasons Intel wishes to remain the single source of the 386 is that IBM of course has an option to fabricate the chip. If IBM decides to pick up that option – and the bug in the original mask is just the kind of irritation that will spur it to take the thing into its own hands – supply would be less of a problem and prices would drop. And IBM has the right not only to make the chip as is, but also to alter the microcode – and you can damn well bet that they will, probably within the next couple of years, says one party. If there are any winners in the current and future 80386 high-price-low-quantity cycle, say some hardware manufacturers, it will probably be the software developers. With fewer 386s coming out in the short term, the opportunity exists to improve the quality of the software, one developer declares. It remains to be seen if the industry responds this way.
Just like a 386
Another manufacturer added that he doubts that software developers are sitting still just because the hardware isn’t there. All parties are in agreement, however, that the short-term losers will be the 80386 clonemakers, who, as one person puts it, have to get things out there now and cheap. Clonemakers will have a hard time. In the long term however, the US chipmakers seem to be in a no-win situation. Motorola has decided that it will ignore pre-existing second-source agreements and keep the 68020 and 68030 to itself, but the policy is likely to be seen to have backfired in a year or two. Because one of those jilted second sources, Hitachi Ltd, has designed its own compatible superset of the 68020, and it is likely to come to market at about the same time as the 68030. Although unlikely to be fully compatible with that device, it will be as compatible with the 68020 as the 68030 is, so that it will almost certainly provide the likes of Sun and Apollo with all they can get from the 68030. If Motorola has any trouble with the new part, or attempts to keep prices up, Hitachi will get its entree to the market. And one hardware manufacturer impatiently awaiting deliveries of 80386s, reckons that with its sole-source policy, Intel is exposing itself to a similar threat. If they delay long enough, he says, they may not be the only player when the market is really huge. We know of several companies that are working on chips that act just like a 386 but aren’t, and Intel won’t be able to claim any infringement. Intel may have gone too far this time in banking on an established market.