Some interesting comments re the speed yields on the Itanium chip from Jim Seymour @ TSC.com.
____________________________________________________________
A Dirty Little Semi Secret
8/23/00 10:17 AM ET
A footnote to the 733MHz-vs.-800MHz speeds on Itaniums:
When INTC announced this week that it would be delivering mainly 733MHz parts, rather than the targeted 800MHz versions, that worried smart observers about Intel's the possible production problems building the chip.
People outside the industry generally don't know the way this works: An Intel produces, in this case, *all* 800MHz parts. Or at least, that's what they're trying to do: no one TRIES to make, say, 733MHz parts.
But in testing the finished "800MHz" chips, some percentage inevitably fail at the designated speed. So they get re-tested at a slightly lower speed, and if they pass there, get relabelled and sold as, in this case, 733MHz parts.
(Are buyers gettings seconds? No, the semi companies, say, because they're getting exactly what the label silk-screened onto the chip says....)
If a semiconductur company produces an imbalance of the two speeds on a given desaign, it sometimes means their production process is flawed, unable to produce most of its wafers and chips that work at the speed they were spec'd to work at.
And of course, overall this problem reduces revenues, since so many parts have to be sold at the substantially-cheaper prices of the slower parts.
Just a bit of Silicon Arcana from the RealMoney.com vaults.... |
