Nvidia Chip Problems Might Be Warning for Everyone

(MoneyWatch)  The Inquirer ran a story about a new wave of Nvidia chip problems. Companies with semiconductor-based products might shake their heads sadly (with a smirk on competitors' faces), but what's happening with Nvidia should make everyone else think twice.

I have to throw a big caveat in here: there is a lot of controversy over Charlie Demerjian'scoverage of Nvidia. If the story and information is accurate, this could be devastating to the company:

It is hard to overstate how bad this is. Basically every 65nm and 55nm Nvidia part appears to be defective. It is not a question of yes or no, but how defective each line is, and what the failure rate for each one is. We are hearing of early failure rates in the teens per cent for 8800GTs and far higher for 9600GTs, so this is not a quibble over split hairs.

Even if it is overblown, Nvidia has clearly had product issueswith a $196 million charge "to cover expected customer warranty, repair, return, replacement, and associated costs arising from a weak die/packaging material set in certain versions of its chipsets and GPU products used in notebook systems."

Put this intro perspective. Only recently did Intel move to 45nm and AMD move off 65nm. Nvidia is hardly new at making chips, and it's apparently been wiping out right and left at both 65nm and 55nm. The cumulative evidence suggests that moving to new semiconductor geometries is way more difficult than people have been admitting.

What's really disturbing is the natureof the problems. They aren't issues that show up in the lab, unless you think that Nvidia managment would be foolish enough to ship chips they knew would fail and need replacement. No, this suggests that the industry has hit such a level of complexity and unforgiving physical tolerances that companies simply may be unable to know what the issues will be until they happen in the field.

The other factor is whether a company owns its manufacturing plant or uses someone else's facilities. Intel has its own fab plant, and AMD has denied rumors that it would sell off its own plants and become a design house only.

In contrast, Nvidia contracts out manufacturing, far and away the most common situation. While that may be economically efficient, given the crushing amount of capital investment that fab plants require, it has a down side. Using a contract fab house means working to some degree at arm's length. Maybe success at these technically challenging chip dimensions requires closer integration and control.

Whatever the case may be, Nvidia's had problems, AMD's had problems, and even Intel had its challenges. Just how easy is it going to be for other semiconductor design firms to get the financial efficiencies they need while still having the smaller size chips work? Probably not easy at all, which suggests that the whole semiconductor industry may be headed for the biggest shake-up it has ever faced. Slower transition means a longer time to realize a return on perhaps billions spent on facilities and an expensive hiccough in the longer-term financial health of the chip designers.