
Subject: --> Final Summary of Petition to Intel Regarding Pentium FDIV

Newsgroups: comp.sys.ibm.pc.hardware.chips,comp.sys.ibm.pc.hardware.misc,
comp.sys.ibm.pc.hardware.systems,comp.sys.intel,misc.consumers,
misc.legal.computing,news.admin.misc
Path: fu-berlin.de!math.fu-berlin.de!zib-berlin.de!news.th-darmstadt.de!terra.wiwi.uni-frankfurt.de!zeus.rbi.informatik.uni-frankfurt.de!news.dfn.de!swiss.ans.net!howland.reston.ans.net!ix.netcom.com!netcom.com!noring
From: noring@netcom.com (Jon Noring)
Subject: --> Final Summary of Petition to Intel Regarding Pentium FDIV
Message-ID: <noringD2IGLs.6oK@netcom.com>
Followup-To: comp.sys.intel
Summary: 2360 Validated Electronic Signatures!
Organization: Netcom Online Communications Services (408-241-9760 login: guest)
Date: Mon, 16 Jan 1995 18:27:28 GMT
Lines: 195
Xref: fu-berlin.de comp.sys.ibm.pc.hardware.chips:19264 comp.sys.ibm.pc.hardware.misc:19978 comp.sys.ibm.pc.hardware.systems:13499 comp.sys.intel:34554 misc.consumers:77914 misc.legal.computing:10941 news.admin.misc:29958



[Note the followup to comp.sys.intel]

Hello,

This post is to summarize the results of the petition to Intel regarding
the Pentium floating divide flaw.  It was started on December 14, 1994,
and official signature acceptance ended yesterday, January 15, 1995.  The
Petition statement is attached as an appendix to this post.

I received 2360 validated and unique signatures, which is quite impressive
considering that Intel relented only a few days after the Petition was
started and essentially agreed to the first two of our three demands.  I'm
certain that had Intel not relented, the number of signatures would have been
much higher, maybe as many as 5000 or more, since signature submission took
an abrupt nose dive within hours of Intel's announcement.  Out of these 2360
signatures, 1008 claimed to own or work on a Pentium-based pc, 977 said they
did not, and 375 did not release this information.  I urge Intel to e-mail
those who own or work on a Pentium pc to make sure their needs are being met
in a timely fashion.

You can download the signature list (an ASCII file) via anonymous ftp from

ftp.netcom.com  /pub/no/noring/pentium.signatures

    [ Ed.Note-17/1/95-khd:  This ASCII file could be found on the
                            following ftp server too.

                            ftp.grumed.fu-berlin.de

                            in directory "PC" as file

                            INTEL_PETITION_SIG.TXT   ]

If you have difficulty retrieving this list, e-mail me and I'll send it to
you via e-mail.  The raw file containing the original e-mail submissions
(which is almost 3 megs in size) will be made available via anonymous ftp on
request.  Intel may be interested in the raw file since it contains many
comments.


The third demand, that Intel publicly release information on how to tell a
FDIV flawed chip from a fixed one, both internally and externally, has not
yet been met.  Intel, do the right thing and release this information!

It's really not proven if the Petition had a substantial impact on Intel in
making their decision to implement a "no questions asked policy."  Private
feedback from my sources who supposedly have connections inside Intel has
been little more than anecdotal, but does suggest it may have pushed Intel
over the edge since they were being assaulted on many fronts.  If anybody
has more solid information on this, we'd all like to know.

Anyway, I thank everybody for their signatures for this very successful
Internet petition drive!  It shows the power of the Internet of bringing
people together for a common purpose.

Jon Noring

(p.s., since I don't have an e-mail address of the right person(s) at Intel
to mail the signatures to, I'd appreciate it if either somebody gives me an
e-mail address or if someone at Intel with the proper authority can download
the information and then let me know that they received it.  I am forwarding
the information to several media contacts I have.  Feel free to forward this
announcement to others.)


****************Beginning of Petition Statement to Intel********************

   Date:  December 14, 1994
Subject:  Petition to Intel on the Pentium Design Flaw

Dear Intel:

We, the undersigned, are very concerned about your handling of the floating
point divide (FDIV) design flaw found in all of your Pentium processors
manufactured until very recently.  We do acknowledge that perfection is hard
to achieve in any product.  However, there are certain types of imperfections
that are serious enough to warrant special attention and treatment.  This is
clearly one of them.

Your actions to downplay the FDIV design flaw have only created confusion and
hostility among your loyal customers.  In addition, the fact that you knew
about the design flaw from July, but did nothing to warn Pentium users about
it, and even continued to ship Pentium processors with the FDIV design flaw
to resellers until very recently (and probably still are), only adds fuel to
the hostility felt by Pentium owners and would-be owners.  We know it is not
your intent to alienate your customers, but that is effectively what you have
done by your mishandling of the situation up to the present.  We urge you to
change course immediately.

You are right in saying that a large number of Pentium users will not be
adversely affected by the FDIV design flaw since they are not using their
Pentiums *now* for intensive floating point calculations.  However, a
significant and also large number of Pentium users, including scientists,
engineers, designers, financial professionals, students, mathematicians and
statisticians, and ordinary spreadsheet users, to name just a few of the
affected groups of Pentium users, could encounter this error, with many
possible negative consequences.  Engineers typically do design calculations
where an error could seriously affect public safety;  financial analysts
dealing with large sums of money where an error in the 4th significant figure
(as the design flaw is sometimes capable of producing) can result in millions
of dollars of losses.  And do not forget that many of your customers bought a
Pentium for what it potentially could do, not just to run applications they
are currently using.

Your claims that it will occur so rarely (e.g., once every 27,000 years for
the typical user) as to be inconsequential are totally unfounded, as many
recent posts by competent scientists and mathematicians to the Usenet/Internet
newsgroup 'comp.sys.intel' have demonstrated beyond a doubt.  In addition,
IBM, a major seller of Pentium systems, recently issued a press release also
doubting your claims, stating that even the ordinary spreadsheet user could
encounter this FDIV error every few days -- for a large company with many
spreadsheet users, this could lead to several errors per day, some of which
could prove costly.  Many scientists and engineers have reported that they
had to throw away months of costly calculations since they are not sure the
answers are correct, or that others will call into question the results
during the peer review/publication process.  Even several Ph.D. students have
reported on Usenet that they have to rerun months of calculations, delaying
the completion of their thesis by several months or more.

In addition, your implied argument that it wasn't discovered for a long time
as being proof that it is not serious, is also incorrect.  As an arithmetic
flaw, which does not give any warning, it is especially insidious because
when it does occur, the user is usually unaware of it (that's why it took so
long to be discovered *and* publicized).  Only by rerunning the code on a
machine with a proven reliable FPU can the error be seen.  Most of these
codes take so long to run, even on the fast Pentium, that rerunning the codes
for verification is very costly and thus are usually not undertaken.  Upon
hearing of the FDIV design, many Pentium users have rerun FPU intensive
calculations on FPU reliable processors and some have reported finding serious
discrepancies that went previously unnoticed.

Your present policy of downplaying the seriousness of the FDIV design flaw,
by using incorrect statistical analysis (your so-called "White Paper"), and
by your massive PR media blitz which tends to only give your side of the
story (you call it a "subtle flaw" with shades of double-speak in George
Orwell's "1984"), is especially serious and adversely affects your
credibility.  Your offer to replace the Pentium processor for those who, in
your opinion, could be affected by the flaw, is tainted by your fairly strict
and seemingly arbitrary qualification criteria, as well as your lackluster
effort (effectively none) in trying to inform Pentium owners about the FDIV
design flaw.  Your so-called "Six Point Plan" to handle the situation is
clearly motivated by damage control rather than by trying to satisfy your
customers.  There is a BIG difference.

In summary, your mishandling of the entire situation only serves to alienate
both your customers and stockholders, and may even make you legally liable
for endangering the public safety and for economic losses of various sorts.
We are certain you and your stockholders don't want this.

We, the undersigned, therefore make the following demands, which we hope you
will consider as friendly recommendations from your loyal customers:

1) That you will offer a free replacement Pentium processor (which has been
   updated so as to fix the FDIV design flaw without slowing down FPU
   performance) to any owner of a FDIV-flawed Pentium processor who requests
   a replacement, with *no questions asked*, and with *no exceptions*.  In
   addition, you will make it easy, logistically, for the requester to obtain
   the replacement processor, and you will pay all reasonable costs for
   effecting the replacement.

2) That you will make a *good faith* effort to inform all Pentium owners of
   your replacement offer in demand #1 above, for example, by half to
   full-page advertising in major newspapers, magazines, trade and scientific
   journals, notices to Pentium system resellers, universities, research
   laboratories, etc.  All notices and advertisements must include a clearly
   objective and concise assessment, written by an expert in the field not
   associated *in any way* with Intel, of the FDIV design flaw and in what
   type of applications it could cause problems.  The assessment must
   reference, *without comment*, information recently released by IBM
   concerning the potential frequency of FDIV errors to spreadsheet users.
   The notices and advertisements must also include the information from
   demand #3 below.  The campaign to inform Pentium owners must begin within
   six weeks from the acceptance of our demands.

3) That you will publicly release information on how to determine, by the
   serial number printed on the processor, a FDIV design flawed Pentium from
   one that has this design flaw fixed.  This information must be posted to
   the Internet as well as provided to your resellers and to the editors of
   the major PC-related consumer magazines and PC-industry newsletters.
   

We, the undersigned, believe the above demands are not harsh, and will help
you to gain credibility and confidence among all computer users.  In fact,
this may actually result in a marketing victory for you, as people speak
enthusiastically and do business with companies who follow the principle that
"the customer is always right."  They will see you as a company who cares for
the little guy, with the result that your long-term success is assured.
Without such consumer confidence, no business has ever been able to thrive in
the long-term.  We ask that you choose to put your customers first, and the
long-term profits will surely follow.  We are certain that your stockholders
would also wholeheartedly agree with this assessment.

Intel, do the right thing!

Signed,

********************End of Petition Statement to Intel***********************


-- 
OmniMedia           | The Electronic Bookstore.  Come in and browse!  Two
1312 Carlton Place  | locations:  ftp.netcom.com  /pub/Om/OmniMedia/books
Livermore, CA 94550 | and  ftp.awa.com  /pub/softlock/pc/products/OmniMedia
510-294-8153        | E-book publishing service follows NWU recommendations.  


