Four years ago: Why Intel wouldn’t withdraw the lot

PR disaster equivalent to San Francisco fire of 1904

While Intel acknowledged that the muckup over the Pentium FPU bug was a PR disaster equivalent to the San Francisco fire of 1904 last week, there are very good reasons why it won't simply swap the lot out. A total recall would be dangerous for the entire PC industry, according to an insider at Intel last week. Intel still maintains it is on target to ship between six to seven million Pentiums by the end of the week, a figure hotly disputed by other racehorse owners in the Desperation Derby. Although it can undoubtedly afford the cost of replacing the existing parts, this would mean delays of several months, a stall in sales and people like Gateway, Dell and Packard Bell stopping selling Pentium machines. (A source at Cyrix, incidentally, reckoned last week that Dell and Gateway were paying more for their Pentiums anyway than industrial distributors). Intel's thinking is that the stall in the marketplace could potentially put the big Pentium-champions out of business, a move which would cause less competition, and slow down technology development. Journalists funded by advertising would also lose their jobs, Intel thinks. Interestingly, few of Intel's competitors have made hay out of the problems with the Pentium. They know that pride comes before a fall and the same problem could happen to them. System manufacturers, the insider said, should carry on shipping product and work with customers. He said that the whole affair was like a whoopee cushion. Intel now acknowledges that Andy Grove's posting in COMP.SYS.INTEL was a PR mistake. A press conference would have been better. The advice given to other players in the industry by Intel seems to have taken...in the last week we received a flurry of press releases from software, hardware and networking manufacturers attempting to calm the situation right down. These range from Compaq through Apricot to Microsoft and Lotus. Did Intel accidentally plan it? Contemplating his corporate plan that seemed to indicate substantially greater persistence for the 486 than Intel has been anticipating for most of this year, a grumpy Compaq source told us that back in Q1 Compaq had looked at Intel's projections and told the company its prices were all wrong. Compaq planned accordingly, Intel wound up cutting prices to more acceptable levels through 94, but the net effect was that the bigger, planning-intensive companies shipped a lot fewer Pentiums, percentage-wise, than the young upstarts who made a virtue of turning on a sixpence (old British coin - we don't use them now either). So when the bug showed up, most of the old guard didn't have that many machines out there, and IBM in particular could present itself as the consumer's friend for the bargain price of $2.5 million (100,000 machines at $25 a pop), while Gateway and Dell could go... The total cost of replacing every chip shipped would be $400-600 million in labour, but the collateral damage of tracking machines, getting engineers out to them, fixing them is obviously far greater. Particularly for lean, mean companies like, er Gateway and Dell. And what if it hadn't been fixed? Yes, OK, so we didn't cover this story when it first came out, and now we're beating it to death. But it was a trivial, no-hoper story to start with, it's the sound of grinding axes that have made it big. Anyway, Intel has been pretty shifty on the existence of a fix. The stuff it has been releasing majors on resolving users' concerns, not on giving them a new CPU. Now, the company now concedes it has a fixed unit, and that it should have all the users happy well before the end of Q1. Under some duress, Intel folks will concede that the fix is just coming out of the production process now, indicating that they've had it for two-to-three months. So isn't it odd they didn't say this earlier? And as they've known about it since the Summer, one can imagine a desperate period of several months when Intel was trying to nail it before the solids hit the fan, and failing. Now, if it had taken another two months, how screwed would Intel have been? ® From The Register No. 10

Sponsored: Driving business with continuous operational intelligence