Advocacy
Dojo (HowTo)
Reference
Markets
Museum
News
Other
Myths
Press
General
Hack
Hardware
Interface
Software
Standards
People
Forensics
Web
CodeNames
Easter Eggs
History
Innovation
Sightings
Opinion
Martial Arts
ITIL
Thought
|
In mid '94 Intel (and independant parties) found a floating point bug in their processor. While it is not unusual for Hardware bugs in processors it is rare that they are this serious, furthermore Intel went out of their way to hide the bug. It was about 6 months later that Intel finally went public with the bug (1). Even then Intel did not valuntarily go public with the bug, it was because other people were discovering it and a thread started on the internet.
Dr. Thomas Nicely was the first guy to publicly discover the bug, and he contacted Intel Tech Support regarding this bug on Monday 24 October (call reference number 51270). But he was soon silenced by Intel with an NDA as Intel tried to bury the problem. (2) The bug was found on many different pentiums including the 60, 66, 75 and 90 mhz pentium chips, and effected millions of users.
After some cajoling Intel admitted that the bug existed - but they told people that only professionals that could prove they needed the bug fixed would get their processors replaced (3). Users had to prove to intel that they needed the fixed chip to get it replaced. Intel minimized the bug by distorting how serious it was to many people telling people "It only happens to calculations at the 8th place" or that it was only an issue to researchers and scientists.
Here are some samples that look a lot more serious than the 8th or 9th place -
- 5505001/294911 = 18.66600093 (pentium)
- 5505001/294911 = 18.666651973 (powerpc)
- Notice - the 4th decimal place.
- X = 5505001
- Y = 294911
- Z = (X/Y)*Y - X
- --------------
- Expected: 0
- Computed with a pentium with the bug: -256.00000
- X = 4195835
- Y = 2.9999991
- Z = (X/Y)*Y - X
- --------------
- Expected: 0
- Computed with a pentium with the bug: -192.00000
Intel also implied that the bug only effected FDIV (floating point division), but it was later proved that many functions used the FDIV tables and could also be effected. A few transidental functions effected included the following; FPTAN, FPATAN, FYL2X, FYL2XP1, FSIN, FCOS, FSINCOS. (floating point tangent, sins, cosines, exponents).
After some class action law suits got started, and negative publicity was spinning out of control, Intel agreed to replace faulty chips unconditionally.
Even the resolution cost PC users lots of money (4). How would you like to find out that your machine cost you $300 more than you expected, not to mention thousands or millions in liabilty because of faulty calculations? That was a reality that most early pentium users faced.
The issue was never as much the bug itself. The bug was critical to many users, and converted many users computers from a cost saving device into a liability. The bigger issue was that Intel handled the situation very poorly, by first hiding the facts, then trying to minimize them, and not owning up to their responsibility and requiring users to prove to Intel that they needed replacement. (5) There are also still many pentium computers out there that users never got around to replace. So next time you see an "Intel Inside" sticker, you might want to think of it as a warning label.
For more fun try - Humor
Pentium
Bug Humor
SUPPORTING ARTICLES
(1) "Some Scientists Are Angry Over Flaw in Pentium Chip, and Intel's Response" ~ Wall Street Journal (11/25/94) P. B4; Clark, Don
- Some scientists are angry about a flaw that Intel revealed in its Pentium chip that could make computers answer incorrectly in certain rare instances. The defect, which is in millions of microprocessors, involves division problems with the floating-point processor. The problem was discovered early last summer [1994], the company says, and the production processes were changed accordingly. An Intel spokesman says, "The chip is fine. Statistically, the average person might see this problem once in every 27,000 years." But mathematicians working in precise fields such as chaos theory take the flaw very seriously. Researcher Dave Bell says, "There are a lot of people who do research and have to stand up and publish their results based on computer simulations. Maybe one of the questions now will be, was it done on a chip with the bug or without the bug?"
(2) Response from Dr. Nicely- FROM: Dr. Thomas R. Nicely, Professor of Mathematics, Lynchburg
- College, Lynchburg, Virginia ([email protected])
- TO: Whom It May Concern
- RE: Pentium FPU
- DATE: 16 November 1994
- This is in response to the many e-mail inquiries I have recently received. I have entered into a temporary nondisclosure agreement with Intel Corporation, and am therefore not at liberty to comment further on the Pentium CPU and FPU.
- Information regarding the Pentium CPU and FPU may be obtained by calling Intel directly at 800-628-8686, or by contacting Intel representative John Thompson at 408-765-1279.
(3) "A Flaw Chips Away at Intel's Shiny Image" ~ Washington Post (12/02/94) P. A1; Corcoran, Elizabeth
- Intel has lost some of its reputation with the revelation that its Pentium chip is flawed. Not only has Intel's reputation for quality been compromised, but analysts say its handling of the situation has been very poor. Intel CEO Andrew Grove has issued an apology over the Internet, but the company says a mass recall is unnecessary because the flaw becomes apparent with only the most complex calculations. Intel will sell its existing inventory of flawed chips until it runs out sometime next year. The company has reserved the right to decide if a customer needs a replacement chip.
(4) "The Real Cost" ~ InformationWeek (01/09/95) No. 509, P. 12; Gillooly, Brian
- Intel's decision to provide Pentium replacements to all customers upon request may end up costing large companies thousands of dollars. Although Intel replaces the chip for free, companies must handle reconfiguration, testing, customer downtime, and equipment depreciation. Analyst Carter Lusher says replacing Pentiums may cost companies an average of $289 per system, including administrative, labor, and downtime costs. Smith Barney in New York will have to replace Pentiums in 600 PCs, which could cost the company over $200,000. Analyst Liz Buyer says, "A charge of $200,000 would certainly turn some heads.
(5) "A Flaw Chips Away at Intel's Shiny Image" ~ Washington Post (12/02/94) P. A1; Corcoran, Elizabeth
- Intel has lost some of its reputation with the revelation that its Pentium chip is flawed. Not only has Intel's reputation for quality been compromised, but analysts say its handling of the situation has been very poor. Intel CEO Andrew Grove has issued an apology over the Internet, but the company says a mass recall is unnecessary because the flaw becomes apparent with only the most complex calculations. Intel will sell its existing inventory of flawed chips until it runs out sometime next year. The company has reserved the right to decide if a customer needs a replacement chip. Analyst Drew Peck says, "Clearly [the problem] was grossly mishandled from a public relations standpoint. Analyst Erik Jansen says, "I can't believe that Intel is trying to tell a customer we will decide what makes you happy. They've successfully created awareness of Pentium. Now Pentium means watch out."
|
