CNN Interviews Thomas Nicely
David R. Conrad
On CNN's WorldNews at about 10:30 pm EST on December 12th they aired
an interview with Prof. Thomas Nicely regarding the FDIV error in the
Pentium. When promoing the interview they said that IBM has decided
to stop "shipments of computers that include the flawed Intel Pentium
chip. IBM says problems with the chip are more serious than
previously believed. The widely used Pentium chip miscalculates
certain sophisticated math problems. But Intel says the average user
should not be affected. After IBM's announcement Monday, Intel stock
dropped more than $2 a share." At another point it was the
"problematic Pentium chip" which "miscalculates certain math problems"
and the statement that Intel "has offered to replace the chip only for
those customers who can prove they use their PCs for complex math
equations."
I taped the interview between CNN's Bobbie Battista and Prof. Nicely,
and have made this transcript of it. It is reprinted here without
permission, as "fair use".
-=-=-
CNN: Well, that now notorious glitch in the Pentium computer chip was
first discovered by a mathematics professor at Lynchburg College in
Virginia and Professor Thomas Nicely joins us now by telephone.
Professor, thanks for joining us.
Nicely: Glad to join you.
CNN: How did you happen to discover this problem?
Nicely: I was working on a research problem in pure mathematics; it
involved prime numbers and a huge number of divisions had to be
performed over a long period of time and in the process of the
calculation I discovered at one point that there was a discrepancy
and it took several months to track it down and it turned out to be
the least likely suspect of all: the chip.
CNN: Oh. Well there seems to be quite a discrepancy now about how
damaging this is. I mean, IBM is claiming that, you know, you could
run into an error maybe as often as once a month and others are saying
maybe once in 27,000 years.
Nicely: Yeah, I'm very skeptical of the statistics I've heard
attributed to IBM today. The claims I've heard are that they think
a spreadsheet user, the average spreadsheet user, might encounter an
error once every 24 days, and at the same time they say they accept
Intel's statistics on the statistical frequency of the error. Those
things are simply inconsistent; I would not expect the average
spreadsheet user to see an error once every 50 years, for that matter.
I would tend to agree with much more with Intel's statistics than with
IBM's.
CNN: Were you surprised, then, at IBM's move?
Nicely: Yes, I was very surprised. I was not necessarily so surprised
that they would stop shipment temporarily, but the rational they gave
I found astonishing.
CNN: Do you think there's any other motivation behind it?
Nicely: Well, if I were the suspicious type I would be wondering about
the fact that IBM doesn't produce Pentium chips and Intel does, and
one of the competitors for the Pentium chip is IBM's PowerPC chip
which might stand to gain an advantage if the Pentium chip were
downgraded.
CNN: Is there any way, if you are one of those people that work with
serious math problems and you have this kind of computer work, is
there any way that you can determine within your computer whether or
not you have the faulty chip?
Nicely: Oh yes, there're some fairly simple and short codes you can
download off the Internet to check for the error, and in fact all you
have to do is try one specific division problem and it will determine
whether or not the error's in your machine.
CNN: Okay, and I would assume Intel can tell you about that.
Nicely: Yes, that's correct.
CNN: All right. Professor Nicely, thanks very much for joining us.
Nicely: Thank you.
-=-=-
He hesitated before giving the "50 years" figure in such a way that made
it obvious that it was 'off the top of his head', so to speak. I note
this not to make any point, but only because it isn't obvious from the
transcript alone.
Regards,
David R. Conrad
--
ab...@detroit.freenet.org / David_...@mts.cc.wayne.edu
"No his mind is not for rent to any god or government." -- Rush
Ron Newman
David R. Conrad <ab...@detroit.freenet.org> wrote:
>CNN: Oh. Well there seems to be quite a discrepancy now about how
>damaging this is. I mean, IBM is claiming that, you know, you could
>run into an error maybe as often as once a month and others are saying
>maybe once in 27,000 years.
>
>Nicely: Yeah, I'm very skeptical of the statistics I've heard
>attributed to IBM today.....
Let's give Professor Nicely a break here. It's obvious that CNN
called him before he had the opportunity to read IBM's analysis,
forcing him to comment off the top of his head.
--
Ron Newman MIT Media Laboratory
rne...@media.mit.edu
Don Anderson
Organization: Eskimo North (206) For-Ever
In article <1994Dec13.1...@news.media.mit.edu>
rne...@media.mit.edu (Ron Newman) writes:
>In article <3cjqh3$l...@detroit.freenet.org>,
>David R. Conrad <ab...@detroit.freenet.org> wrote:
>
>>damaging this is. I mean, IBM is claiming that, you know, you could
>>run into an error maybe as often as once a month and others are saying
>>maybe once in 27,000 years.
>>
>>Nicely: Yeah, I'm very skeptical of the statistics I've heard
>
[...]
>--
>Ron Newman MIT Media Laboratory
If, as has been reported on this group, Dr. Nicely has been retained as
a consultant for Intel, neither CNN nor Dr. Nicely thought to mention
this during the interview.
Big Don <big...@eskimo.com>
mpa...@axp2.umkc.edu
>>damaging this is. I mean, IBM is claiming that, you know, you could
>>run into an error maybe as often as once a month and others are saying
>>maybe once in 27,000 years.
>>
>>Nicely: Yeah, I'm very skeptical of the statistics I've heard
>>attributed to IBM today.....
>
> Let's give Professor Nicely a break here. It's obvious that CNN
> called him before he had the opportunity to read IBM's analysis,
> forcing him to comment off the top of his head.
It's obvious that Intel is paying Prof. Nicely, and, being on Intel's payroll,
he is reluctant to say anything that could make Intel look dishonest.
Michael Passer
Ken Lui
Ron Newman <rne...@media.mit.edu> wrote:
> It's obvious that CNN
>called him before he had the opportunity to read IBM's analysis,
>forcing him to comment off the top of his head.
If this were truely the case, Professor Nicely should not have
commented about IBM's analysis in the first place.
Ken
--
Ken Lui, kl...@corp.hp.com 3000 Hanover Street MS 20CG
Palo Alto ITC Palo Alto, CA 94304-1112 USA
Client/Server Engineering 1.415.857.3230 FAX 1.415.857.5518
Clifton T. Sharp
>called him before he had the opportunity to read IBM's analysis,
>forcing him to comment off the top of his head.
Still, there's something ironic about his instantaneous change from
27,000 years to 50 years...
--
Cliff Sharp Never get into fights with ugly people.
WA9PDM They have nothing to lose.
cli...@indep1.chi.il.us --The Fourth Law of Reality
103t_e...@west.cscwc.pima.edu
> In article <1994Dec13.1...@news.media.mit.edu>,
> Ron Newman <rne...@media.mit.edu> wrote:
>> It's obvious that CNN
>>called him before he had the opportunity to read IBM's analysis,
>>forcing him to comment off the top of his head.
>
> If this were truely the case, Professor Nicely should not have
> commented about IBM's analysis in the first place.
>
>
I understand that Prof. Nicely is seriously ill and possibly having a major
operation soon. THat CNN could get him to come to the phone at all may have
been a major coup.
Lawson
John Navas
No need for that -- he was even more disparaging of IBM in the San
Francisco Examiner article published yesterday, and again failed to
mention any relationship with Intel.
--
Best regards and happy holidays,
John <JNa...@Netcom.com>