Digital's lost opportunity: the RISC flip-flop - Digital Equipment Corp.'s wavering focus on development of reduced-instruction-set-computer chips

Was last month's acquisition of MIPS Computer by Silicon Graphics the last chapter in the story of RLSC at Digital? Unless Alpha becomes a huge hit (with lots of OEM customers and great success in PC markets), it may not be.

Several years ago Digital was working on a project called Prism to develop RISC chips in house. But, in September of 1988, Digital shelved Prism in favor of an OEM agreement with MKPS, citing a six-month time-to-market advantage the company would realize by using the outsourced chips. Flip.

Abandoning Prism had deleterious short-term side effects. The most important was that Digital couldn't hold on to some of the brightest RISC engineers in the business. To wit, Prism veteran David Cutler quickly went to Microsoft where he now heads NT development.

Apparently, however, Digital didn't completely discard its in-house RISC projects. While outwardly the company sold MIPS, it was not entirely committed to it. The company continued quietly developing what is now Alpha. During the past year, when rumors of MIPS demise were rampant (after major OEMS like Bull soured on KUPS), Digital might even have purchased MIPS, had the internal commitment been there. It wasn't. Flop. During that period, Digital was never truly committed to a single RISC architecture.

Last month, Digital officially announced what has been assumed, unofficially, for more than a year: internal RISC is still alive at Digital with Alpha. Flip again.

The net result is that Digital's flip-flops have muddied its hardware strategy, from PCs to supercomputers. It boasts two RISC architectures, its competitors one. Hewlett-Packard, IBM and Sun, made a choice and stuck to it That's why they have the advantage (in single-architecture RISC deliverables) over Digital today.

The late start does not necessarily mean Digital is sunk. Digital is in the same position HP was several years ago. Only its product-fine transition may be a little less painful because it is moving from 32- to 64-bit machines as opposed to HP going from 16- to 64-bit.

On the other hand, history is not in Digital's favor. In the early 1980s three companies introduced microprocessors into the commodity chip market. Intel was first. Its 8088 family has gone on to unrivalled success. The product line succeeded not because of amazing technology, but because Intel got to market "firstest with the mostest."

Motorola was second. Its 68000 was a little faster and a little more elegant than Intel's chips. Despite their technical edge, the Moto chips have remained behind Intel in terms of market share, revenues and overall profitability.

National Semiconductor's 16000 family was the latecomer. It was faster yet, and even more powerful than the 68000. But the 16000 family never really succeeded. Its advantages were rendered less significant because the market had coalesced around the two earlier products.

Alpha is no 16000. National Semi never had a base like the VAX from which to build. And it could never have even pretended that the 16000 might have a 25-year lifespan. In commodity microprocessors, the early bird gets the worm. Digital is hoping it can leapfrog today's RISC leaders and focus on the next-generation repast.

COPYRIGHT 1992 International Data Corporation
COPYRIGHT 2004 Gale Group