des@yatton.inmos.co.uk (David Shepherd) (10/30/89)
Friday's (27oct89) Financial Times had an article headlined "Compaq shares drop as it finds `problem' with chip" Brief summary: Compaq announced that it had found problem with fpu on an Intel chip it is using in its new generation PCs. An analyst is quoted as saying "You could just hear an audible gasp in the crowd when they made the announcement". The result of this was a 10% fall in the value of Compaq stocks and 5% fall in Intel stocks. Neither company was clarifying what the problem or its implications actually were. Anyone got any more info on what's happenning .... I assume its the 80486 at fault. david shepherd INMOS ltd ps Compaq just seem to be unlucky when it comes to Intel fpus .... they got very bored with waiting for correct 80387s as well ;-)
ok@cs.mu.oz.au (Richard O'Keefe) (10/31/89)
In article <2725@ganymede.inmos.co.uk>, des@yatton.inmos.co.uk (David Shepherd) writes: > Brief summary: Compaq announced that it had found problem with fpu on > an Intel chip it is using in its new generation PCs. An analyst is > quoted as saying "You could just hear an audible gasp in the crowd when > they made the announcement". The result of this was a 10% fall in the > value of Compaq stocks and 5% fall in Intel stocks. Given that Compaq *found* the problem and were honest enough to talk about it, in a just world their share prices should have *risen*.
dmocsny@uceng.UC.EDU (daniel mocsny) (10/31/89)
Warning: purely frivolous post follows. Technically inclined readers with important work to do please hit 'n' now. > In article <2725@ganymede.inmos.co.uk>, des@yatton.inmos.co.uk (David Shepherd) writes: > > Brief summary: Compaq announced that it had found problem with fpu on > > an Intel chip it is using in its new generation PCs. An analyst is > > quoted as saying "You could just hear an audible gasp in the crowd when > > they made the announcement". The result of this was a 10% fall in the > > value of Compaq stocks and 5% fall in Intel stocks. With a stock movement of this magnitude, the armchair conspiratorialist in me has to wonder whether anyone connected with the fpu bug might have been in position to profit. While I am no expert in stock futures, I do know that by taking the appropriate market position, a speculator can wager large sums on various stock price movements. A very clean conspiracy scenario is where an engineer surreptitiously plants a crippling, yet subtle bug that is sure to slip through preliminary testing. When the bug surfaces late enough in the game to blow holes in a major product introduction, the resulting stock price perturbation delivers a handsome profit to the speculators who earlier gambled that the price would drop. With a company with revenues the size of a Compaq, we could be talking about enough money to provide someone with a comfortable early retirement. However, some careful laundering would be necessary, and if the firm knew who was responsible for the bug, (s)he would have a most difficult time enjoying the loot anytime soon. Also, if the bug got caught and fixed too early, the speculators would take a sizable bath, and might do something very unpleasant to their co-conspirator as a result. Disclaimer: I have no connection with any of the above firms, except as a rabidly satisfied customer, and neither do I have the slightest shred of evidence for any of the above statements. The above is purely speculation of the most pathetic armchair variety, and no doubt is riddled through with factual errors and conceptual distortions. Furthermore, I am vaguely aware of the strict ethical standards that all the organizations and their personnel doubtless uphold to the letter, which would prevent even the thought of such a heinous scheme from ever occurring. (I sleep better thinking that.) And of course my employer would instantly pretend to not know me if any of this got back to them... :-) But it would make a great movie... Dan Mocsny dmocsny@uceng.uc.edu
dolf@idca.tds.PHILIPS.nl (Dolf Grunbauer) (10/31/89)
>In article <2725@ganymede.inmos.co.uk>, des@yatton.inmos.co.uk (David Shepherd) writes: > Brief summary: Compaq announced that it had found problem with fpu on > an Intel chip it is using in its new generation PCs. What is the actual problem ? What chips 80486 and/or 80487 ? Could someone give more details or did they only say there is a problem and not what the precise problem is ? -- Dolf Grunbauer Tel: +31 55 433233 Internet dolf@idca.tds.philips.nl Philips Telecommunication and Data Systems UUCP ....!mcvax!philapd!dolf Dept. SSP, P.O. Box 245, 7300 AE Apeldoorn, The Netherlands --> Holland is only 1/6 of the Netherlands <--
ron@woan.austin.ibm.com (Ronald S. Woan) (10/31/89)
In article <443@ssp1.idca.tds.philips.nl>, dolf@idca.tds.PHILIPS. Grunbauer) writes: > What is the actual problem ? What chips 80486 and/or 80487 ? > Could someone give more details or did they only say there is a problem > and not what the precise problem is ? According, second hand, to the San Jose Mercury news 27 Oct 89, a defect causes incorrect calculations given a rare sequence of floating point operations. Ghastly! Ron +------All Views Expressed Are My Own And Are Not Necessarily Shared By-------+ +-------------------------------My Employer-----------------------------------+ + Ronald S. Woan (IBM VNET)WOAN AT AUSTIN, (AUSTIN)ron@woan.austin.ibm.com + + outside of IBM @cs.utexas.edu:ibmaus!auschs!woan.austin.ibm.com!ron +
terrell@cadnetix.COM (Eric Terrell) (11/01/89)
What was the reaction to the announcement of the 32 bit multiply error in the 386? Terrell
mslater@cup.portal.com (Michael Z Slater) (11/01/89)
The bug Compaq referred to was in the FPU on the 486. It occurs only under very specific and rare conditions. Intel says only that a customer found it, and won't identify which one; Compaq says it was them. An indication of how obscure the bus is is that the 486 passes all of Intel's test suites for the 387. Intel says that they have engineered the fix, and will have corrected parts by the end of November. Michael Slater, Microprocessor Report mslater@cup.portal.com
ed@imuse.intel.com (Ed Braaten) (11/02/89)
In article <443@ssp1.idca.tds.philips.nl> dolf@idca.tds.PHILIPS.nl (Dolf Grunbauer) writes: >What is the actual problem ? What chips 80486 and/or 80487 ? What is an 80487??? Do you know something I don't know? Ed ------------------------------------------------------------------------- Ed Braaten | "Love the Lord your God with all your Internet: ed@imuse.intel.com | heart and with all your soul and with EUnet: unido!imuse!ed | all your strength." Deuteronomy 6:5 -------------------------------------------------------------------------
davidsen@crdos1.crd.ge.COM (Wm E Davidsen Jr) (11/02/89)
In article <10136@cadnetix.COM>, terrell@cadnetix.COM (Eric Terrell) writes: | | What was the reaction to the announcement of the 32 bit multiply | error in the 386? I don't think there's a correlation. The 386 (at the time the multiply bug was found) was being primarily used for DOS, and most applications didn't use the 32 bit instructions. The problem with the might be found in any o/s, at least as described in the magazines. Personally, *if* Intel is telling us the truth, I see no problem. They claim to have a mask revision, they say they will have it in place by the end of the month, and they say they can still ship production quantities this year. While I would not hesitatte to use one of the old 386's in a 16 bit environment (with the machine properly labeled of course) I would not use one of the flawed 486's for anything. I sort of hope I can persuade Intel to give me one, even burned out if they wish, for a big hi-tech tie tack ;-) -- bill davidsen (davidsen@crdos1.crd.GE.COM -or- uunet!crdgw1!crdos1!davidsen) "The world is filled with fools. They blindly follow their so-called 'reason' in the face of the church and common sense. Any fool can see that the world is flat!" - anon
conor@inmos.co.uk (Conor O'Neill) (11/03/89)
Reproduced from Electronics Weekly, Wed, November 1, 1989, front page: (without permission) Two design faults have been discovered in Intel's million transistor 80486 microprocessor chip which was first unveiled last April and has been in production since August. Both faults are in the chip's floating point unit: one affects its ability to correctly calculate trigonometrical functions such as tangents, cosines and sines; while the other (in the exception handling detection system) affects the floating point unit's ability to detect whether a particular calculation makes sense. "We've already shipped thousands of chips and yes if people ask for replacements we will do it," said an Intel spokeswoman. Intel has redesigned the chip, made new masks and is now manufacturing a new version of the 80486 which is free of these two bugs, she added. [end of article] As usual, not enough information for anyone to use. -- Conor O'Neill, Software Group, INMOS Ltd., UK. UK: conor@inmos.co.uk US: conor@inmos.com "It's state-of-the-art" "But it doesn't work!" "That is the state-of-the-art".
joel@nosun.UUCP (Joel Clark) (11/04/89)
In article <443@ssp1.idca.tds.philips.nl> dolf@idca.tds.PHILIPS.nl (Dolf Grunbauer) writes: >>In article <2725@ganymede.inmos.co.uk>, des@yatton.inmos.co.uk (David Shepherd) writes: >> Brief summary: Compaq announced that it had found problem with fpu on >> an Intel chip it is using in its new generation PCs. > >What is the actual problem ? What chips 80486 and/or 80487 ? >Could someone give more details or did they only say there is a problem >and not what the precise problem is ? >-- >Dolf Grunbauer Tel: +31 55 433233 Internet dolf@idca.tds.philips.nl >Philips Telecommunication and Data Systems UUCP ....!mcvax!philapd!dolf >Dept. SSP, P.O. Box 245, 7300 AE Apeldoorn, The Netherlands > --> Holland is only 1/6 of the Netherlands <-- From DATAQUEST information services. 80846 CHIP FIX At its quarterly financial analyst meeting held on October 26, Intel announced that it was officially releasing availability information on its 80486 32-bit microprocessor, on which an error resulting from a floating point operation had been discovered. Dave House, general manager of the microcomponents group essentially described the problem as "obscure". The error affects floating point tangent and exception handling and arises when the floating point instruction is handling exact powers of 2 and when the instruction is followed by a particular sequence of instructions. Readers may recall that in its implementation of the 80486, Intel had incorporated the floating point unit on-chip. Discovery of the error result was made after 100,000 hours of testing performed by Intel and some of its major customers who had received early shipment of the device. Once Intel had verified the problem at the chip level, the company proceeded to notify customers on Friday, October 20th. Intel is in the process of implementing a fix on the mask set and expects to have evaluation product from the revised masks next week. In the meantime, shipments of the 80486 are on-hold. Many wafers in production had not yet reached the final metalization stage, and so are being held prior to metalization pending the new masks. Production will resume in November and the company still expects to ship tens of thousands of the 80486 during the fourth quarter. Intel policy with respect to the flawed devices is to offer its customers replacements. Intel will continue to supply the defective 80486 to customers wishing to go ahead and use the device. Many business applications will not be affected by the floating point error. The floating point tangent operation is a new instruction not featured on the original 8087 math-coprocessor. Moreover, some workarounds have been developed to handle the resulting floating point error which may arise in scientific and technical applications. Resolution of this problem will not impact Intel as severely as the problem the company had with the 80386, largely because the problem was discovered at an early stage. In the case of the math errors occurring on the 80386 were discovered over a year after shipments had commenced. Shipments of the 80486 had only begun early in the third quarter, and most customers have not yet released their products based on the chip. Therefore, the number of devices affected is smaller and substitution will be simplified by the relatively low extent of distribution. Intel stated that this problem would not affect anticipated fourth quarter financials, which had previously been characterized as providing modest revenue growth. joel clark intel scientific computers joel@intelisc.intel.com (503) 629-7732
c9h@psuecl.bitnet (11/07/89)
In article <2800@ganymede.inmos.co.uk>, conor@inmos.co.uk (Conor O'Neill) writes: > "We've already shipped thousands of chips and yes if people ask for > replacements we will do it," said an Intel spokeswoman. Intel has > redesigned the chip, made new masks and is now manufacturing a new version > of the 80486 which is free of these two bugs, she added. > > [end of article] > > As usual, not enough information for anyone to use. Not enough information? What the Hell do you want? They offered to replace the chips, for Christ's sake! Jesus! Some people are never satisfied! ('scuse me, but stupidity really irks me!) -- - Charles Martin Hannum II "Klein bottle for sale... Inquire within." (and PROUD OF IT!!!) "To life immortal!" c9h@psuecl.psu.edu "No noozzzz izzz netzzznoozzzzz..." cmh117@psuvm.psu.edu "Memories, all alone in the moonlight ..."
toms@omews44.intel.com (Tom Shott) (11/07/89)
There us no such chip as the i487. The functionality of the i387 was improved and integrated into the i486. Some improved functionality resulted from a tighter integration of the FPU into the processor. (How faults and exceptions are handled.) Floating point performance was improved 3-4X compared to a i387 due to reduced number of cycles for some operations. Disclamer: I DO work for Intel. I had nothing to do w/ the i486. The information stated is from the Intel i486 data sheet (book). -- ----------------------------------------------------------------------------- Tom Shott INTeL, 2111 NE 25th Ave., Hillsboro, OR 97123, (503) 696-4520 toms@omews44.intel.com OR toms%omews44.intel.com@csnet.relay.com INTeL.. Designers of the 960 Superscalar uP and other uP's
kenobi%lightsabre@Sun.COM (Rick Kwan) (11/07/89)
In article <2800@ganymede.inmos.co.uk> conor@inmos.co.uk (Conor O'Neill) writes: >Reproduced from Electronics Weekly, Wed, November 1, 1989, front page: >(without permission) > >Two design faults have been discovered in Intel's million transistor 80486 >microprocessor chip which was first unveiled last April and has been in >production since August. > >Both faults are in the chip's floating point unit: one affects its ability >to correctly calculate trigonometrical functions such as tangents, cosines Anyone know: 1. how long (man-years and/or calendar years) it took to write the test vectors to exercise the 486, and roughly how many vectors are involved? (I heard that the 386 took 3 years.) 2. what kind of system was used for testing (Trillium, Sentry, or what)? I assume this is an absolutely full-custom chip. Yes? Rick Kwan Sun Microsystems - Intercontinental Operations kenobi@sun.UUCP or kenobi%lightsabre@sun.COM
scott@bbxsda.UUCP (Scott Amspoker) (11/08/89)
In article <63979@psuecl.bitnet> c9h@psuecl.bitnet writes: >In article <2800@ganymede.inmos.co.uk>, conor@inmos.co.uk (Conor O'Neill) writes: >> [Intel announces that mysterious 486 bug has been fixed] >> >> As usual, not enough information for anyone to use. > >Not enough information? What the Hell do you want? They offered to replace >the chips, for Christ's sake! Jesus! Some people are never satisfied! I agree with Conor. I want to know exactly what the problem was. It is not enough that Intel will ship new 486 chips. When I have a customer on the phone complaining about a bug in our software I want to recognize the symptoms of an early 486 that somehow did not get updated. -- Scott Amspoker Basis International, Albuquerque, NM (505) 345-5232 unmvax.cs.unm.edu!bbx!bbxsda!scott
andrew@dtg.nsc.com (Lord Snooty @ The Giant Poisoned Electric Head ) (11/12/89)
I believe the Weitek 4167 (?) boosts 80486 fpt performance by "a factor of 3 or 4". In any case, how does the coupled 80486/4167 system decide which fpt unit will execute a given fpt instruction? and is this a way around the bug, temporarily - in that the Weitek would take over the sin/cos/tan stuff? -- ........................................................................... Andrew Palfreyman a wet bird never flies at night time sucks andrew@dtg.nsc.com there are always two sides to a broken window
darryl@ism780c.isc.com (Darryl Richman) (11/12/89)
In article <291@berlioz.nsc.com> andrew@dtg.nsc.com (Lord Snooty @ The Giant Poisoned Electric Head ) writes:
"I believe the Weitek 4167 (?) boosts 80486 fpt performance by "a factor of
"3 or 4". In any case, how does the coupled 80486/4167 system decide which
"fpt unit will execute a given fpt instruction? and is this a way around the
"bug, temporarily - in that the Weitek would take over the sin/cos/tan stuff?
The Weitek is memory mapped device and works on a completely different
instruction set than the x87. You can actually configure asystem and get
all three processors running in parallel. To use a Weitek, you need to
compile for it in advance, and if you don't have one (or an emulator),
such a binary is useless.
--Darryl Richman
--
Copyright (c) 1989 Darryl Richman The views expressed are the author's alone
darryl@ism780c.isc.com INTERACTIVE Systems Corp.-A Kodak Company
"For every problem, there is a solution that is simple, elegant, and wrong."
-- H. L. Mencken