Ethics in Engineering Case 01:
The Intel Pentium® Chip
Engineers must have an understanding of technical factors involved in making professional
and ethical decisions. Here’s a brief exercise:
The Intel Pentium® Chip
In late 1994, the media began to report that there was a flaw in the new Pentium
microprocessor produced by Intel. The microprocessor is the heart of a personal computer
and controls all of the operations and calculations that take place. A flaw in the Pentium
was especially significant since it was the microprocessor used in 80% of the personal
computers produced in the world at that time.
Apparently, flaws in a complicated integrated circuit such as the Pentium, which at the time
contained over one million transistors, are common. However, most of the flaws are
undetectable by the user and don’t affect the operation of the computer. Many of these
flaws are easily compensated for through software. The flaw that came to light in 1994 was
different: It was detectable by the user. This particular flaw was in the floating-point unit
(FPU) and caused a wrong answer when double-precision arithmetic, a very common
operation, was performed.
A standard test was widely published to determine whether a user’s microprocessor was
flawed. Using spreadsheet software, the user was to take the number 4,195,835, multiply it
by 3,145,727, and then divide that result by 3,145,727. As we all know from elementary
math, when a number is multiplied and then divided by the same number, the result should
be the original number. In this example, the result was 4,195,579 [Infoworld, 1994].
Depending on the application, this six-thousandths-of-a-percent error might be very
significant.
At first, Intel’s response to these reports was to deny that there was any problem with the
chip. When it became clear that this assertion was not accurate, Intel switched its policy
and stated that although there was indeed a defect in the chip, it was insignificant and the
vast majority of users would never even notice it. The chip would be replaced for free only
for users who could demonstrate that they needed an unflawed version of the chip
[Infoworld, 1994]. There is some logic to this policy from Intel’s point of view, since over
two million computers had already been sold with the defective chip.
Of course, this approach didn’t satisfy most Pentium owners. After all, how can you predict
whether you might have a future application where this flaw might be significant? IBM, a
major Pentium user, canceled the sales of all IBM computers containing the flawed chip.
Finally, after much negative publicity in the popular personal computer literature and an
outcry from Pentium users, Intel agreed to replace the flawed chip with an unflawed
version for any customer who asked to have it replaced.
It should be noted that long before news of the flaw surfaced in the popular press, Intel was
aware of the problem and had already corrected it on subsequent versions. It did, however,
continue to sell the flawed version, and, based on its early insistence that the flaw did not
present a significant problem to users, seemingly planned to do so until the new version
was available and the stocks of the flawed one were exhausted. Eventually, the damage
caused by this case was fixed as the media reports of the problem died down and as
customers were able to get unflawed chips into their computers. Ultimately, Intel had a
write-off of 475 million dollars to solve this problem.
What did Intel learn from this experience? The early designs for new chips continue to have
flaws, and sometimes these flaws are not detected until the product is already in use by
consumers. However, Intel’s approach to these problems has changed. It now seems to feel
that problems need to be fixed immediately. In addition, the decision is now based on the
consumer’s perception of the significance of the flaw, rather than on Intel’s opinion of its
significance.
Indeed, similar flaws were found in 1997 in the early versions of the Pentium II and
Pentium Pro processors. This time, Intel immediately confirmed that the flaw existed and
offered customers software that would correct it. Other companies also seem to have
benefited from Intel’s experience. For example, Intuit, a leading manufacturer of tax
preparation and financial software, called a news conference in March of 1995 to apologize
for flaws in its TurboTax software that had become apparent earlier in that year. In
addition to the apology, they offered consumers replacements for the defective software.
