My Perspective on Pentium - AGS
Richard Wirt
weekend and we are out of the office, I am posting from my home system.
Richard Wirt
Director SW Technology
Intel Corp
This is Andy Grove, president of Intel. I'd like to comment a bit on
the conversations that have been taking place here.
First of all, I am truly sorry for the anxiety created among you by
our floating point issue. I read thru some of the postings and it's
clear that many of you have done a lot of work around it and
that some of you are very angry at us.
Let me give you my perspective on what has happened here.
The Pentium processor was introduced into the market in May of '93
after the most extensive testing program we at Intel have ever
embarked on. Because this chip is three times as complex as the 486,
and because it includes a number of improved floating point
algorithms, we geared up to do an array of tests, validation, and
verification that far exceeded anything we had ever done. So did many
of our OEM customers. We held the introduction of the chip several
months in order to give them more time to check out the chip and their
systems. We worked extensively with many software companies to this
end as well.
We were very pleased with the result. We ramped the processor faster
than any other in our history and encountered no significant problems
in the user community. Not that the chip was perfect; no chip ever
is. From time to time, we gathered up what problems we found and put
into production a new "stepping" -- a new set of masks that
incorporated whatever we corrected. Stepping N was better than
stepping N minus 1, which was better than stepping N minus 2. After
almost 25 years in the microprocessor business, I have come to the the
conclusion that no microprocessor is ever perfect; they just come
closer to perfection with each stepping. In the life of a typical
microprocessor, we go thru half a dozen or more such steppings.
Then, in the summer of '94, in the process of further testing (which
continued thru all this time and continues today), we came upon the
floating point error. We were puzzled as to why neither we nor anyone
else had encountered this earlier. We started a separate project,
including mathematicians and scientists who work for us in areas other
than the Pentium processor group to examine the nature of the problem
and its impact.
This group concluded after months of work that (1) an error is only
likely to occur at a frequency of the order of once in nine billion
random floating point divides, and that (2) this many divides in all
the programs they evaluated (which included many scientific
programs) would require elapsed times of use that would be longer than
the mean time to failure of the physical computer subsystems. In
other words, the error rate a user might see due to the floating point
problem would be swamped by other known computer failure mechanisms.
This explained why nobody -- not us, not our OEM customers, not the
software vendors we worked with and not the many individual users --
had run into it.
As some of you may recall, we had encountered thornier problems with
early versions of the 386 and 486, so we breathed a sigh of relief
that with the Pentium processor we had found what turned out to be a
problem of far lesser magnitude. We then incorporated the fix into
the next stepping of both the 60 and 66 and the 75/90/100 MHz Pentium
processor along with whatever else we were correcting in that next
stepping.
Then, last month Professor Nicely posted his observations about this
problem and the hubbub started. Interestingly, I understand from
press reports that Prof. Nicely was attempting to show that
Pentium-based computers can do the jobs of big time supercomputers in
numbers analyses. Many of you who posted comments are evidently also
involved in pretty heavy duty mathematical work.
That gets us to the present time and what we do about all this.
We would like to find all users of the Pentium processor who are
engaged in work involving heavy duty scientific/floating point
calculations and resolve their problem in the most appropriate fashion
including, if necessary, by replacing their chips with new ones. We
don't know how to set precise rules on this so we decided to do it
thru individual discussions between each of you and a technically
trained Intel person. We set up 800# lines for that purpose. It is
going to take us time to work thru the calls we are getting, but we
will work thru them. I would like to ask for your patience here.
Meanwhile, please don't be concerned that the passing of time will
deprive you of the opportunity to get your problem resolved -- we
will stand behind these chips for the life of your computer.
Sorry to be so long-winded -- and again please accept my apologies
for the situation. We appreciate your interest in the Pentium
processor, and we remain dedicated to bringing it as close to
perfection as possible.
I will monitor your communications in the future -- forgive me if I
can't answer each of you individually.
Andy Grove
Rolfe Tessem
>Andy Grove has asked me to post the following for him. Since it is the
>weekend and we are out of the office, I am posting from my home system.
(much deleted -- BTW, a posting coming from an intel.com address would
have a little more credibility; don't you have the technology to
access your office network remotely? If not, I'm sure Livingston has
some products they would love to sell you -- but I digress...).
>Sorry to be so long-winded -- and again please accept my apologies
>for the situation. We appreciate your interest in the Pentium
>processor, and we remain dedicated to bringing it as close to
>perfection as possible.
Not long-winded enough, in my opinion. Missing is a real "white-paper"
detailing the actual extent of the problem. Intel has a
public-relations disaster on it's hands, and the longer the company
resists coming forward with full-disclosure, the worse it will get.
Are you guys in some kind of denial or what? I suggest getting in
touch with Johnson & Johnson and talking to them about how they
handled the Tylenol poisoning and its aftermath.
In the meantime, what are customers who are about to order
Pentium-based products supposed to do? I was going to order an HP
Netserver LM this week, pricing out at about $18,000.00. In the
absence of a statement from Intel that they'll replace processors in
new machines at least, I'll be forced to buy something else since
this machine has to be up and running by Jan. 1. The machine will be
running an accounting application that we base our entire business on;
what would you do?
--
Rolfe Tessem Lucky Duck Productions
ro...@ldp.com 96 Morton Street
ro...@ibm.net New York, NY 10014
Bill Seward
>
>Andy Grove has asked me to post the following for him. Since it is the
>weekend and we are out of the office, I am posting from my home system.
>
>Richard Wirt
>Director SW Technology
>Intel Corp
OK, making the unwarrented assumption that this post is legit...
>
>
>This is Andy Grove, president of Intel. I'd like to comment a bit on
>the conversations that have been taking place here.
>
>First of all, I am truly sorry for the anxiety created among you by
>our floating point issue. I read thru some of the postings and it's
>clear that many of you have done a lot of work around it and
>that some of you are very angry at us.
"very angry"? Incensed might be a more accurate term. For some of
us, there is the small but real possibility that Bad Things have happened
or could happen as a result of this bug--esoteric or not.
<much stuff about chip design and fab deleted, along with notes of bugs
that have occured in previous Intel CPUs>
>That gets us to the present time and what we do about all this.
>
>We would like to find all users of the Pentium processor who are
>engaged in work involving heavy duty scientific/floating point
>calculations and resolve their problem in the most appropriate fashion
>including, if necessary, by replacing their chips with new ones. We
>don't know how to set precise rules on this so we decided to do it
>thru individual discussions between each of you and a technically
>trained Intel person. We set up 800# lines for that purpose. It is
>going to take us time to work thru the calls we are getting, but we
>will work thru them. I would like to ask for your patience here.
Sir, this simply will not do. While my Pentium is not currently used
for "heavy duty scientific/floating point calculations", I can in no
way guarentee that this will hold true forever. While the future plans
for this particular machine are for it to be migrated to a WWW server,
I also have several folks who would like to have it (or one like it)
for their desks--including one user who is dying to run SAS for Windows
on it.
< more deleted>
>I will monitor your communications in the future -- forgive me if I
>can't answer each of you individually.
>
>Andy Grove
>
I hope you do. I have already stated publicly that I will not specify
any more Pentium-based machines for purchase at my employer (who shall
remain nameless) until our supplier (DEC) can guarentee me that they are
shipping machines with the bug-free chip. And I am pretty much the final
word on this subject, being the LAN manager. I hate to buy 486s, but I
will.
While I do not recall the names of the other folks, there have been
posts that others are going to do likewise.
In addition, my planned Christmas present purchase of a new p-90 for my
family's personal use has been tabled. For some time, I have considered
the new Power Macs as a viable alternative to the Intel/Windows platform.
Barring a major movement by Intel on this issue by New Year's, I will
probably move in this direction. And with some regret, because up until
now I have been a solely Intel kinda guy.
Sir, I hope that you and your firm will "do the right thing" and provide
replacement chips for all who ask for them. While I'm sure the cost
would be significant to Intel, the lost good will will cost you more in
the long run.
Bill Seward
w...@infi.net
I speak for myself. No one else will.
Franklin Antonio
Richard Wirt <rw...@ix.netcom.com> wrote:
>Andy Grove has asked me to post the following for him. Since it is the
>weekend and we are out of the office, I am posting from my home system.
Thank you for posting this. Hearing the Intel position first-hand
really helps.
>... We then incorporated the fix into
>the next stepping of both the 60 and 66 and the 75/90/100 MHz Pentium
>processor along with ...
I have a couple of questions about the corrected (ie no FDIV bug) chips to
which I would appreciate an official intel answer.
Are the corrected chips available now?
When did Intel begin shipping them in quantity?
Mike Palij
>
>Andy Grove has asked me to post the following for him. Since it is the
>weekend and we are out of the office, I am posting from my home system.
>
>Director SW Technology
>Intel Corp
>
>
>the conversations that have been taking place here.
>
>First of all, I am truly sorry for the anxiety created among you by
>our floating point issue. I read thru some of the postings and it's
>clear that many of you have done a lot of work around it and
>that some of you are very angry at us.
As the unhappy recipient of three brand new buggy Gateway
Pentia I was hoping that the rest of Andy Grove's alleged
message might have gone something like this:
We appreciate the fact that you can't predict the situations
in which the Penitum bug is likely to occur, especially since
the nature of your work may change in short notice or others
may need to use your Pentium for jobs which may invoke the
error. Frankly, we screwed up and our currently stated policy
is extremely short-sighted and works to the disadvantage of
the most important person in this situation, namely you, the
customer. We've decided that the only reasonable solution
to this situation is to replace buggy chips for anyone who
wants them replaced. No questions asked. Please call the
following 800 number to start the return process....
If the real Andy Grove does read this, all I can say is that
there still is time for you to do the right thing.
**************=Ain't No One's Opinion But My Own=**************
- Michael Palij -
Psychology Dept. - pa...@xp.psych.nyu.edu - New York University
Robert Ashcroft
Please pass the following onto Mr Grove, assuming that the
message below, posted to the Internet on Sunday, is actually from
the Intel chief executive.
I am about to purchase a Pentium system on behalf of my father, a
science professor at an Ivy League school. We have a quote from
a well-known vendor (Hi-Q) and are ready to purchase this week.
We will require a statement in writing from Hi-Q that they
guarantee the proposed system will be free of the Pentium FDIV
bug. Should the vendor be unable to provide such a guarantee,
we will buy another product, possibly DEC Alpha.
Further, should the product arrive with the Pentium bug nonetheless,
we will return it as DOA.
It is nothing short of outrageous that Intel presumes to judge whether
its customers really need the Pentium performance Intel promised in
the first place. Financially, Intel can afford to replace all bad
chips. Ethically and reputationally, Intel cannot afford not to, and
moreover, Intel has done so in the past, with early 386 processors.
Stay true to precedent and repair the damage to your reputation while
you still can.
Robert Ashcroft
In article <3bamq9$a...@ixnews1.ix.netcom.com>,
Richard Wirt <rw...@ix.netcom.com> wrote:
Luc Bauwens
>In <3bamq9$a...@ixnews1.ix.netcom.com>, rw...@ix.netcom.com (Richard Wirt) writes:
>>Andy Grove has asked me to post the following for him. Since it is the
>>weekend and we are out of the office, I am posting from my home system.
>
>(much deleted -- BTW, a posting coming from an intel.com address would
>have a little more credibility; don't you have the technology to
>access your office network remotely? If not, I'm sure Livingston has
>some products they would love to sell you -- but I digress...).
>
>>Sorry to be so long-winded -- and again please accept my apologies
>>for the situation. We appreciate your interest in the Pentium
>>processor, and we remain dedicated to bringing it as close to
>>perfection as possible.
>
>Not long-winded enough, in my opinion. Missing is a real "white-paper"
>detailing the actual extent of the problem. Intel has a
>public-relations disaster on it's hands, and the longer the company
>resists coming forward with full-disclosure, the worse it will get.
>Are you guys in some kind of denial or what? I suggest getting in
>touch with Johnson & Johnson and talking to them about how they
>handled the Tylenol poisoning and its aftermath.
But way too long winded regarding all that BS on how great they are
as a company on how great the Pentium is. That was downright offensive.
All that letter achieved was:
1. to make the case that the way Intel handled the affair was endorsed
from the very top;
2. they still don't recognize they are wrong and unethical;
3. they still see this as primarily a PR issue;
4. their policy remains unchanged;
5. Intel does not care about the substance of the issues that we have
been ventilating here, and which his letter doesn't even mention;
6. Mr. Grove will no longer be able to find a scapegoat.
Luc
___
Friends don't let friends buy buggy Pentiums
sic...@pupgg.princeton.edu
>In <3bamq9$a...@ixnews1.ix.netcom.com>, rw...@ix.netcom.com (Richard Wirt) writes:
>>Andy Grove has asked me to post the following for him. Since it is the
>>for the situation. We appreciate your interest in the Pentium
>>processor, and we remain dedicated to bringing it as close to
>>perfection as possible.
>
>In the meantime, what are customers who are about to order
>Pentium-based products supposed to do? I was going to order an HP
>Netserver LM this week, pricing out at about $18,000.00. In the
>absence of a statement from Intel that they'll replace processors in
>new machines at least, I'll be forced to buy something else since
>this machine has to be up and running by Jan. 1. The machine will be
>running an accounting application that we base our entire business on;
>what would you do?
>
application you should not be afflicted by the pentium bug.
you can cerify this with the software company that is selling you
the software, but I don't think that ANY accounting application
uses double precision divides.
Besides since you haven't bought the system yet,
you can DEMAND from your supplier that he guarantees
a FDIV bug-free chip, since they are supposedly available.
Hugues Sicotte
No SIG file!!!
U36...@uicvm.uic.edu
>
>Andy Grove has asked me to post the following for him. Since it is the
>weekend and we are out of the office, I am posting from my home system.
>
>Richard Wirt
>Director SW Technology
>Intel Corp
>
>
>This is Andy Grove, president of Intel. I'd like to comment a bit on
>the conversations that have been taking place here.
>
>First of all, I am truly sorry for the anxiety created among you by
>our floating point issue. I read thru some of the postings and it's
[snip]
I am truly sorry that five weeks ago I bought a Pentium computer
for scientific and engineering work. I am now worried that I am
going to be stuck with a system whose results will always be suspect.
I assure you that had this come up during the 30 day return window this
system would be on its way back to the manufacturer by now. As it
is I intend to bring as much pressure to bear on the manufacturer
under the one year warranty to get my money back or a good P90 chip.
YOUR policies as related in this group DO NOT give me confidence.
>
>This group concluded after months of work that (1) an error is only
>likely to occur at a frequency of the order of once in nine billion
>random floating point divides, and that (2) this many divides in all
>the programs they evaluated (which included many scientific
>programs) would require elapsed times of use that would be longer than
>the mean time to failure of the physical computer subsystems. In
Interesting. The problem is that there is no reason that the error
should not occur earlier rather than later. This is a typical
misuse of probability methods. The fact that I have a greater chance
of being struck by lightning than of winning the lottery does _not_
mean that I have to be struck by lightning before I can win the lottery.
>other words, the error rate a user might see due to the floating point
>problem would be swamped by other known computer failure mechanisms.
>This explained why nobody -- not us, not our OEM customers, not the
>software vendors we worked with and not the many individual users --
>had run into it.
How many users ran into it and didn't know it ?
This is very interesting. It seems to me that you are looking at this
problem in the aggregate of lots of Pentium systems doing lots of
different things, and viewed in that manner it may indeed seem
like a small problem - only a few (relativley) people on only a few
( r.. ) systems.
You *must* realize however, that Pentium owners and users are seeing
this _very_ differently. Some of us are doing important and
sophisticated work. Whether YOU see that work as being sophisticated
or not - we obviously thought that our work was important and
sophisticated enough to buy a state-of-the-art computer on which to
do it. Indeed, Intel marketing led us to believe that the Pentium was
an appropriate platform for sophisticated technical work. Now we find
out that in some simple calculcations (the type that may be embedded
many times in more complex operations) we can expect no better than
four digits of precision. The fact that, statistically, these errors
occur relatively rarely does not reassure us. What we have lost is
the confidence that we had in the quality and the precision of our
tools.
How many of us would have purchased Pentium systems had we been
aware of this problem ? I wouldn't have. I value the confidence
that I have in the quality of my tools. Perhaps this is why you at
Intel kept this problem from us. This is what makes me truly very
angry. In an interview you said that the "fault" is minor,
yet you didn't have the courage of your convictions to disclose
this fault to potential technical users to let them decide if it
might have a deleterious effect upon their current or future
work for themselves. Instead, you made that decision for us.
Now, you propose to make another decision for us.
You propose to have your own technicians review our work and
our requirements, and if we pass this "oral exam" you will
then give us what we originally paid for. Is it any wonder
that we are outraged ?
As for "other computer failure mechanisms", they are not nearly
as insideous as the FDIV problem. Some failures halt the system -
some even lose data - but most are catastrophic in that the failure
is obvious to the user. The FDIV error gives no error message or
other overt indication of its effect. From the standpoint of someone
like myself, it would be better if the error were gross since
then I would have some confidence in being able to catch it.
Also, most other errors are transient or sporadic in nature - one
may effect a calculation once, but subsequent instances of the same
calculation can be error free. The FDIV error is quite consistent -
if the calculation is affected, it will never be correct.
As it stands now, I wish that I'd bought a DX4 rather than the
Pentium, and if I have to return this system you can bet that
I'll take a long hard look at the PowerPC.
Regards,
- Christopher Wieczorek
u36...@uicvm.uic.edu
University of Illinois at Chicago - Engineering Physics
John Woodward
Luc Bauwens <bau...@acs.ucalgary.ca> wrote:
>But way too long winded regarding all that BS on how great they are
>as a company on how great the Pentium is. That was downright offensive.
>
>All that letter achieved was:
>1. to make the case that the way Intel handled the affair was endorsed
>from the very top;
>2. they still don't recognize they are wrong and unethical;
>3. they still see this as primarily a PR issue;
>4. their policy remains unchanged;
>5. Intel does not care about the substance of the issues that we have
>been ventilating here, and which his letter doesn't even mention;
>6. Mr. Grove will no longer be able to find a scapegoat.
>
>Luc
>___
>Friends don't let friends buy buggy Pentiums
Precisely. This entire affair could be handled in a much better way. Now
people like me warn all their faculty associates, friends, and family
that a Pentium is not the way to go right now. Hold off on any purchases
until this is 100% resolved.
Now all I have to do is sit by the phone until some "Software Engineer"
from Intel calls to question my on my software and why I need a Pentium
that functions properly. Oh joy...
john
John Reece
>In <3bamq9$a...@ixnews1.ix.netcom.com>, rw...@ix.netcom.com (Richard Wirt) writes:
>>Andy Grove has asked me to post the following for him. Since it is the
>>weekend and we are out of the office, I am posting from my home system.
>(much deleted -- BTW, a posting coming from an intel.com address would
>have a little more credibility; don't you have the technology to
>access your office network remotely? If not, I'm sure Livingston has
>some products they would love to sell you -- but I digress...).
As of the start of the Thanksgiving holiday Intel's executive staff
was not set up with TCP/IP and USENET newsreaders. So they went
with someone who already had USENET access. Grove reposted
the message Sunday afternoon from an intel.com address, and as of
Monday morning it was still sitting in an incoming spool directory at
our ISP.
John Reece
not an Intel spokesman, except maybe this one time
Rolfe Tessem
Fair enough -- it did eventually show up here, and did seem to have a
long propagation delay. But how about addressing the substantive part
of my comments?
Setting aside for the moment how Intel plans to deal with Pentiums
already in the field, what are new customers supposed to do? We have
an order in with a reseller for an HP Netserver LM. For tax reasons we
need to spend this money before the end of the year, but I cannot let
this order go forward without assurance that the processor will be one
without the FDIV bug. This machine will be used as an application
server for business critical accounting and budgeting (if it were
destined to be a file server, I wouldn't care.) What would you do in
my position?
--
Rolfe Tessem | Lucky Duck Productions
ro...@ldp.com | 96 Morton Street
(212) 463-0029 | New York, NY 10014
tanting_apollo
Woodward) wrote:
For those people who does scientific calculations with their Pentium
machines that are unable to return their computer, I would suggest the next
time they purchase their next machine, they should look at clone
processors. I do not say they are 100% free defective, but you have a
better chance of replacing your defective parts because they have more to
loose than a company who has a total monopolistic control of the PC market.
For instance, next year AMD and Cyrix is releasing their Pentium killer
clones. From what I read from Byte magazine, these Pentium clones(K5 and
M1) is 30% faster than Pentium. Furthermore, the K5 would even outperform
the upcoming P6 from Intel. It is better to wait than jump at the newest
processor. If you could not wait there is other options such as
i486DX4-100. This processor has almost the same performance as the Pentium
60. At least this processor is bug free. Also, there is Sun and DEC.
These workstations should be able to do what Pentium can do but even
better. There is also th AMD 486DX2-80. This processor is at least 10%
faster that i486DX2-66.
"The joy of monopolistic company is they can say and do what they want."
David Gutierrez
Reece) wrote:
> As of the start of the Thanksgiving holiday Intel's executive staff
> was not set up with TCP/IP and USENET newsreaders. So they went
> with someone who already had USENET access.
Maybe they should have contacted one of the intel.com guys that (until
recently, I suspect) are always posting about the superiority of the
Pentium on comp.sys.powerpc and comp.sys.mac.advocacy.
--
David Gutierrez
d...@biomath.mda.uth.tmc.edu
"Only fools are positive." - Moe Howard
Thomas Koenig
article <3bfh2r$n...@daffy.ldp.com>:
>We have
>an order in with a reseller for an HP Netserver LM. For tax reasons we
>need to spend this money before the end of the year, but I cannot let
>this order go forward without assurance that the processor will be one
>without the FDIV bug. This machine will be used as an application
>server for business critical accounting and budgeting (if it were
>destined to be a file server, I wouldn't care.) What would you do in
>my position?
This depends... you have a number of choices; ordering does not
mean preference :-)
- If your application does its arithmetic in BCD, you should be fine
(COBOL or PL/I programs will usually do this).
- If you can get equivalent functionality from another hardware platform,
such as a HP running UNIX or an Alpha running VMS, Windows NT or OSF,
choose that platform. This, of course, depends critically on the
availability of networking you need (TCP/IP? Novell?) and on
the avilability of compilers/programs for these platforms.
- See wether the performance of a 486/DX100 will suffice.
Thomas "can't believe that he's recommending WNT to somebody" Koenig
--
Thomas Koenig, Thomas...@ciw.uni-karlsruhe.de, ig...@dkauni2.bitnet.
The joy of engineering is to find a straight line on a double
logarithmic diagram.
Thomas J. Teixeira
(Richard Wirt) wrote:
> Andy Grove has asked me to post the following for him. Since it is the
> weekend and we are out of the office, I am posting from my home system.
>
> Richard Wirt
> Director SW Technology
> Intel Corp
>
>
> This is Andy Grove, president of Intel. I'd like to comment a bit on
> the conversations that have been taking place here.
[Stuff deleted]
> This group concluded after months of work that (1) an error is only
> likely to occur at a frequency of the order of once in nine billion
> random floating point divides, and that (2) this many divides in all
> the programs they evaluated (which included many scientific
> programs) would require elapsed times of use that would be longer than
> the mean time to failure of the physical computer subsystems. In
> other words, the error rate a user might see due to the floating point
> problem would be swamped by other known computer failure mechanisms.
> This explained why nobody -- not us, not our OEM customers, not the
> software vendors we worked with and not the many individual users --
> had run into it.
Many of these "known computer failure mechanisms" are not repeatable. For
example, double-bit memory errors are not detected by simple memory
parity.
- You can fool all of the people some of the time.
In contrast, the FDIV error is repeatable (FDIV does come up with the same
answer every time you give it the same arguments).
- You can fool some of the people all of the time.
Even so, you can reduce your exposure to randomly occurring,
non-repeatable failures by repeating the calculations ON THE SAME MACHINE.
Moreover, some number of these random failures will cause the entire
machine to fail so the error is at least detectable (you don't get any
answer because the program crashed).
With the current design flaw with FDIV, you can run your program as many
times as you want and still get the same answer, but you are still left
with the problem of determining whether your answer is correct.
Rolfe Tessem
>In article <3bavcd$o...@egremont.ldp.com>, ro...@egremont.ldp.com (Rolfe Tessem) writes:
>>In <3bamq9$a...@ixnews1.ix.netcom.com>, rw...@ix.netcom.com (Richard Wirt) writes:
>>>Andy Grove has asked me to post the following for him. Since it is the
>>deleted stuff
>>>Sorry to be so long-winded -- and again please accept my apologies
>>>for the situation. We appreciate your interest in the Pentium
>>>processor, and we remain dedicated to bringing it as close to
>>>perfection as possible.
>>
>stuff deleted
>>In the meantime, what are customers who are about to order
>>Pentium-based products supposed to do? I was going to order an HP
>>Netserver LM this week, pricing out at about $18,000.00. In the
>>absence of a statement from Intel that they'll replace processors in
>>new machines at least, I'll be forced to buy something else since
>>this machine has to be up and running by Jan. 1. The machine will be
>>running an accounting application that we base our entire business on;
>>what would you do?
>Not to take Intel's side but if you are running an accounting
>application you should not be afflicted by the pentium bug.
>you can cerify this with the software company that is selling you
>the software, but I don't think that ANY accounting application
>uses double precision divides.
I have received similar responses via email from a number of people,
including someone within Intel. If true, this does provide some
comfort. It still gives one pause though -- what if we decide to
juggle platforms and run something else on this machine next year that
does do heavy floating point?
>Besides since you haven't bought the system yet,
>you can DEMAND from your supplier that he guarantees
>a FDIV bug-free chip, since they are supposedly available.
Well, this is what I've done, but it is causing confusion for the
reseller because HP has not announced yet what their policy will be. I
did receive email from someone at HP that said they would move quickly
to establish a policy -- maybe by the end of this week. I assume this
is because HP sells into technical markets and feels their users may
be among the most affected by the FDIV bug.
--
Rolfe Tessem | Lucky Duck Productions
ro...@ldp.com | 96 Morton Street
bpo...@alfo06.attmail.com
In article <3bbtvq$q...@nntp.Stanford.EDU>, <r...@leland.Stanford.EDU>
writes:
> From: r...@leland.Stanford.EDU (Robert Ashcroft)
> Newsgroups: comp.sys.intel
> Subject: Reply to "Andy Grove"
> Date: 28 Nov 1994 06:39:54 GMT
> Dear Mr Wirt,
> Please pass the following onto Mr Grove...
>
> I am about to purchase a Pentium system...
>... Should the vendor be unable to provide such a guarantee,
> we will buy another product, possibly DEC Alpha.
> It is nothing short of outrageous that Intel presumes to judge ...
I 'was' in the market for a Pentium, now I don't know. I am an AT&T
employee, if I deliver 'known' bugs to customers, we lose revenue. So it
should be elsewhere. Sorry, Intel, your level of customer service and
quality control leave something to be desired in this situation.
Instead of buying an Intel based computer, I will wait for a clone or
quite probably go with the DEC Alpha. After all, they are on their 2nd
generation of 64 bit chip, bug free. (Or do we have to say that they are
only 'known' bug free?) Where may I ask, is Intel?
I may be persuaded to buy the Pentium processor, but only if Intel agrees
to replace the faulty processor for ALL customers that ask. There should
be absolutely no reason for Intel to ask me or anyone else what I am
doing with my computer. This is especially true for people involved in
research and development that is proprietary in nature.
Back off Mr. Grove, show some Intel 'goodwill'.
Bill Porter
Rolfe Tessem
>Rolfe Tessem (ro...@ldp.com) wrote in comp.sys.intel,
> article <3bfh2r$n...@daffy.ldp.com>:
>>We have
>>an order in with a reseller for an HP Netserver LM. For tax reasons we
>>need to spend this money before the end of the year, but I cannot let
>>this order go forward without assurance that the processor will be one
>>without the FDIV bug. This machine will be used as an application
>>server for business critical accounting and budgeting (if it were
>>destined to be a file server, I wouldn't care.) What would you do in
>>my position?
>
>This depends... you have a number of choices; ordering does not
>mean preference :-)
>
>- If your application does its arithmetic in BCD, you should be fine
> (COBOL or PL/I programs will usually do this).
The application itself may or may not use floating point -- I'm trying
to clarify this with the vendor. However, analysis of the numbers
generated will certainly use floating point, and, as we know, this
bug can easily be replicated in Excel.
>- If you can get equivalent functionality from another hardware platform,
> such as a HP running UNIX or an Alpha running VMS, Windows NT or OSF,
> choose that platform. This, of course, depends critically on the
> availability of networking you need (TCP/IP? Novell?) and on
> the avilability of compilers/programs for these platforms.
I'm looking into other platforms -- the easiest alternative would be
Sun. However, for whatever reason, the vendor prices their application
differently depending on the hardware platform, as well as the number
of concurrent users. The pricing for Intel platforms is substantially
more favorable than for Sun; I'm trying to find out about NT running on
Alpha, since that may be an option.
>- See wether the performance of a 486/DX100 will suffice.
I'm not aware of any server-class machines running this chip. Of
course, we could just integrate one or two ourselves, but the
DX100 really seems to be positioned as a desktop alternative.
>Thomas "can't believe that he's recommending WNT to somebody" Koenig
Can't believe I'm actually considering it, either.
As far as I know, HP has not made an announcement of their
intentions. So, let me make a public offer --
call it "The FDIV Challenge":
We'll purchase a server from any reputable vendor who will deliver
a 100 Mhz Pentium sans FDIV-bug if it can be delivered before the
end of the year. Said server must be capable of running major NOSes
such as Netware, Unixware, and LanServer. Previous quotes from Dell
and HP were in the $12,000-$18,000 range.
--
Rolfe Tessem Lucky Duck Productions
ro...@ldp.com 96 Morton Street
Joey deVilla
Tylenol fiasco years ago? The J&J case is standard reading in any
business ethics course. For those of you who don't remember: some madman
began going into drugstores and poisoning random bottles of Tylenol.
There were a handful of deaths that resulted. Rather than follow
conventional wisodm and institute a cover-up, J&J recalled all Tylenol
bottles, redesigned the packaging to be tamper-proof and started an ad
campaign that outlined the problem and the steps they were taking to
solve it.
Another example: Al Gore admitted he did pot, and little was mentioned of
it afterwards. Bill Clinton said "I didn't inhale" and got a media flap.
The moral? The truth will set you free.
It may cost more in the short run, but if Intel wishes to keeps its
customer base, it should 'fess up and do as much as humanly possible to
correct the problem. The software fix is a step in the right direction,
but not a complete solution. Some suggestions...
- A Johnson & Johnson-style campaign on TV, newspapers and in
computer magazines admitting the problem and outlining fixes.
- Getting on the horn with compiler manufacturers and working out
fixes so that the user has the option of preventing FDIVs from
appearing in object code.
- Posting programs featuring those "bad" numbers that produce the
FDIV errors so programmers know what to look for (such as the
C division program that's been making the rounds. They should
also post "work-arounds".
- Compiling a list of applications known to suffer from the effects
of the FDIV bug. It may take some time to check, but it's worth it.
Priority should be given to precision-sensitive apps such as SPICE
(circuit emulation) and AutoCAD.
- (the biggie) -- a recall. The previous suggestions are stopgap
fixes that can be implemented quickly, but hardware problems needs
hardware solutions.
Who's with me?
: --
: jo...@nomad.phys.lsu.edu
--
_________________________________________________________________________
Joey deVilla [ 3j...@qlink.queensu.ca ]
Queen's University, Kingston, Ontario, Canada
David Gutierrez
deVilla) wrote:
> correct the problem. The software fix is a step in the right direction,
> but not a complete solution. Some suggestions...
>
> - Getting on the horn with compiler manufacturers and working out
> fixes so that the user has the option of preventing FDIVs from
^^^^^
> appearing in object code.
You misspelled "kludges".