For 19 years I took trains to big office buildings, mainly in Japan, and sat down with engineers who worked for big companies. I don't have a good sense of direction, but I learned how to get to hard-to-pronounce places like Musashi Nakahara.
Nearly every interview was about something different, but almost all followed a similar pattern. Enter a large conference room, a young woman dressed in a uniform brings in a coffee, a public-relations person and several engineers enter, and an hour later . . . time to clear out, on schedule.
Going to visit Glenn Henry at IDT Corp.'s Centaur division, on the north side of Austin, Texas, was a different scene altogether. Glenn spent 22 years at IBM, where he tried to convince management to pursue the X86 design path.
He got lucky when he quit IBM and went to work for Dell Computer, where Michael Dell was an upstart in his mid-20s. Henry emerged with some valuable stock, and he credits Dell with teaching him a Keep It Simple design mentality.
Henry's office at IDT is the antithesis of those Fujitsu/Hitachi/NEC conference rooms. A reporter's notebook rests on top of big color plots of the WinChip 4 layout that is Henry's baby now, spread out over a long table. A receptionist doubles as the PR manager and supplies a chicken-sandwich lunch, which fits between piles of discarded magazines.
But this is serious business, and Henry is, if anything, more intense than most of the interviewees from my Japanese days. Henry's tiny band at Centaur (22 designers, 50 people altogether) must bring in IDT's WinChip4 at a lower price than other X86 competitors, ranging from AMD (with an average selling price of $102 or so), National Semiconductor's Cyrix division (ASP in the $60 range) and Rise Technology (pricing unknown). With a small die size as a primary target (60 mm2 in IDT's 0.18-micron process at the turn of the century) and a $30 to $40 target, keeping a lid on costs is essential. But the company must have visibility and credibility.
The bigger question facing Henry & Co. is profitability. With Intel, AMD, Cyrix and the 90-person Rise Technology all going after the same pot of gold, the level of X86 competition is nerve-wracking.
WinChip 4 is Henry's best effort at a small, simple design that runs fast enough to please most wanting an entry-level machine. But with corporate MIS managers and notebook buyers largely remaining loyal to Intel, making money at the bottom end of the market will be tough. Perhaps Henry's surprisingly small design team will thrive in a market where cheap, fast PCs are as appealing as, well, a good chicken sandwich.
