<Not the execution we expect from Intel.. >
This kind of problem is pretty common in a new product startup.The production test sequence does not catch all the possible marginalities on all production units in a new design. Remember that the production test is a brief test that lasts a few seconds and is performed at high temperature. The test sequence is based upon a theoretical model of most likely failure modes, and a characterization of the first samples. Thus problems of this nature are to be expected at this stage, and have occurred on past Intel product start-ups. With the Internet age we becoming more aware of such hiccupps.
These problems are resolved by a longer or more complex test,and by appropriate design tweaks.
The only question in my mind is how much of an impact it will have on total units shipped due to the lower yield, and a longer more complex test sequence.
This problem does not involve the upper management, and is handled by a combination of manufacturing, test and design engineers. The only management issue is, was the best talent assigned to the Coppermine, or was it diverted to Merced or other projects ? Under the pressure to the maximize Coppermine shipments, did the test engineers ignore any marginalities found during characterization of the samples ? Only the persons very close to testing Coppermine can answer these questions. However Intel typically does not resort to witch-hunts, but turns on a dime, confronts problems, and resolves them. |
