i486 machines
Chris Webster
First of all, about the problem with 486 chips: I recall reading precisely
what the problem was, but I can't remember where. Since it evidently wasn't
on here, it must've been in a magazine. Since the only technical magazine
(only because it's the only computing magazine that I read :) is BYTE, I
assume that it was mentioned in the latest issue. The problem occurs
when a certain instruction is executed after another certain instruction.
If there isn't enough time or other instructions executed between them, it
doesn't work right. (Sorry, but that's as technical as I can be.)
In today's Science Times section of the NEW YORK TIMES, one of the PComputingcolumns was devoted to a review of yesterday's Compaq press conference. (Today
is Nov 7th, for those of you who don't get this on the same day.) They
mentioned the two new Compaq i486 machines, but unfortunately in not great
detail. Just hype; they even quoted someone on the floor as saying "poor
ibm". Evidently the higher end model sits on the floor (sounded to me
like a tower configuration AT) and can have 2 i486's or 2 386's, or
combinations of both combos in it. For $15grand and up, it outperforms
minicomputers. (Any one have reliable technical info?) Evidently, the
problems with the chip have been worked out, since suppossively (according
to the Net) Compaq found the problems.
Disclaimer: I'm only an Undergrad, so my opinions do not reflect those
of Syracuse University. They make too much sense for the administration
to agree with. :-)
Tim S. Boshart
}
} First of all, about the problem with 486 chips: I recall reading precisely
} what the problem was, but I can't remember where. Since it evidently wasn't
} on here, it must've been in a magazine. Since the only technical magazine
} (only because it's the only computing magazine that I read :) is BYTE, I
} assume that it was mentioned in the latest issue. The problem occurs
} when a certain instruction is executed after another certain instruction.
} If there isn't enough time or other instructions executed between them, it
} doesn't work right. (Sorry, but that's as technical as I can be.)
}
} In today's Science Times section of the NEW YORK TIMES, one of the PComputingcolumns was devoted to a review of yesterday's Compaq press conference. (Today
} is Nov 7th, for those of you who don't get this on the same day.) They
} mentioned the two new Compaq i486 machines, but unfortunately in not great
} detail. Just hype; they even quoted someone on the floor as saying "poor
} ibm". Evidently the higher end model sits on the floor (sounded to me
} like a tower configuration AT) and can have 2 i486's or 2 386's, or
} combinations of both combos in it. For $15grand and up, it outperforms
} minicomputers. (Any one have reliable technical info?) Evidently, the
} problems with the chip have been worked out, since suppossively (according
} to the Net) Compaq found the problems.
I was in Chicago yesterday to look at the new machines. Here is some
more info:
The high end model is the new Compaq SystemPro. It is currently
available with a 33MHz 386 chip, and will be available with a 33MHz 486
chip when it is available. The system can be configured with either one
or two processors. Here are some of the specifications from some of the
literature I picked up:
System Processor Flexible support for 33-MHz 386 and 486 processor
Design: technology; support for two system processors.
386/33 System 32-bit 33-MHz 386 microprocessor; 33-MHz
Processor on All 82385 Cashe Memory Controller with 64 Kbytes of
386-based high-speed static RAM; separate sockets for optional
Models: 33-MHz Intel 387 and 33-MHz Weitek 3167 numeric
coprocessors.
Multiprocessor COMPAQ 386/33 System Processor Board;
Options: compatible with Novell NetWare 386, SCO UNIX
System V/386 and COMPAQ LAN MANAGER
386/486.
System COMPAQ Flexible Advanced Systems Architecture
Architecture: with Multiprocessing Support integrates multiple
system processors, a separate 32-bit processor/memory
bus and the 32-bit EISA I/O bus for highest possible
system throughput.
Expansion Slots: Total of 11 expansion slots; seven 8/16/32-bit EISA
and four 32-bit processor/memory; eight slots available
on all models.
Memory: 4 megabytes of 32-bit memory standard, expandable
to 256 megabytes. Shared system memory design with
80-ns enhanced page memory.
Disk Storage Base unit (386-240) comes with a 240MB 2-Drive
Devices: Array, Model 386-420 with a 420MB 2-drive array,
Model 386-840 comes with 840MB 4 drive array. System
can be expanded to 1.5GB internal storage, and a
total of 4.28GB combined internal and external.
--
Tim S. Boshart | uunet!adspp!timsb | "After a time, you may find that
Abacus Data Systems | ti...@adspp.UUCP |'having' is not so pleasing a thing
2707 Middlebury St. | 219-295-4290 (voice)| as 'wanting.' It is not logical,
Elkhart, IN 46516 | 219-522-2964 (fax) | but it is often true." -- Spock
Gerald Hawkins
> First of all, about the problem with 486 chips: I recall reading precisely
> what the problem was, but I can't remember where.
Our newspaper here (San Jose Mercury) carried an article detailing a
problem with multiplies and divides. Intel says the problem will occur
when processing floating point multiplies and divides. The implication
was that one could work around it by emulating. Obviously, that defeats
half the purpose of using the '486.
> In today's Science Times section of the NEW YORK TIMES, one of the PComputingcolumns was devoted to a review of Compaqs 11/6 press conference. They
> mentioned the two new Compaq i486 machines, but unfortunately in not great
> detail.
-
Even having worked at a place where it was possible to release a computer
with a major bug known, I still don't see how you can do it! Why would
any customer buy a PC with a known bug? They'll have to release some dog
version of BIOS which looks for the killer instructions, slowing down all
FP operations.
Comment? Anybody from Compaq out there? Anybody going to buy a new
Compaq '486 machine even knowing the '486 has a floating point bug?
" I dreamed I saw Joe Hill last night, alive as you and me.
Said I, 'But Joe, you're ten years dead.'
'I never died' said he. "
Jerry ( je...@starfish.Convergent.COM )
-----
Stephen M. Dunn
$ what the problem was, but I can't remember where. Since it evidently wasn't
There are actually two problems ... one has to do with executing the FSINCOS
and FTAN (I think those are the correct mnemonics) instructions when your
data meets certain conditions ... the other has to do with trapping errors
in virtual memory mode or something like that.
For more information, I suggest reading through comp.sys.intel or comp.arch.
There has been a fair bit of discussion of these bugs in those two groups.
$ minicomputers. (Any one have reliable technical info?) Evidently, the
$ problems with the chip have been worked out, since suppossively (according
$ to the Net) Compaq found the problems.
According to Intel, the problems have been found and corrected; all
chips shipped from now on are correct, and old chips can be returned for
replacement.
--
Stephen M. Dunn cs4...@maccs.dcss.mcmaster.ca
<std_disclaimer.h> = "\nI'm only an undergraduate!!!\n";
****************************************************************************
They say the best in life is free // but if you don't pay then you don't eat
mcdo...@uxe.cso.uiuc.edu
-
>Even having worked at a place where it was possible to release a computer
>with a major bug known, I still don't see how you can do it! Why would
>any customer buy a PC with a known bug? They'll have to release some dog
>version of BIOS which looks for the killer instructions, slowing down all
>FP operations.
Simple. Some buyers don't know about the bugs. Remember, IBM sells
a lot to corporate accounts. The buyers of these don't know beans
about computers. They just buy what IBM tells them to.
I point out that I had no idea that my Model 80 was defective when I got
it - I only found out when some programs bombed. I admit I hadn't
checked carefully about it before getting it, as it was free.
Look inside any IBM 386 machine - for well over a year the machines
they sold were ALL defective - the 386 itself is bad. Look in your
older 386 box at the 386. IF the chip doesn't say 80386DX, it
is defective (not counting 386SX's, of course). Just TRY to get IBM to fix
it! It CAN be done, but it isn't easy. It took us months of bitching.
Just remember IBM's motto "We don't care!".
Doug McDonald
Wm E Davidsen Jr
| Comment? Anybody from Compaq out there? Anybody going to buy a new
| Compaq '486 machine even knowing the '486 has a floating point bug?
Why not? The bug occurs infrequently and Intel has already said it
will replace the chips on request. I'm waiting for the 33MHz part
mayself, the problems which are too big for a 386 are going to need at
least 2:1 improvement to be practical.
--
bill davidsen (davi...@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
Steven Philip Lin
>Even having worked at a place where it was possible to release a computer
>with a major bug known, I still don't see how you can do it! Why would
>any customer buy a PC with a known bug? They'll have to release some dog
>version of BIOS which looks for the killer instructions, slowing down all
>FP operations.
>
>Comment? Anybody from Compaq out there? Anybody going to buy a new
>Compaq '486 machine even knowing the '486 has a floating point bug?
Well, I'm not from Compaq but I did work at Micronics a few summers ago when
386 computers were just coming out. When the 32-bit multiplication error
was discovered in the 386 there was a big scare that everybody would stop
buying 386's. That never happened; the purchase orders kept on coming in.
This was before Intel had announced their exchange policy (i.e. if you
have a defective chip you can trade it in when we get bug free chips), and
Intel waited two or three months after discovery of the bug before announcing
an exchange back policy. Apparently the fascination for wanting a new product
overcame people's reservations for buying a defective product. The 386 was
still fancy enough to warrant buying, bug or no bug. Plus there was
virtually no software out there that used 32-bit multiplication so the bug
wasn't really all too serious.
From what I hear at Micronics, the same thing is happening with the
486. People still want to buy the new 486 boards, even with the buggy
chip because they know they can just exchange the 486 for a new one when Intel
ships the fixed chips.
Russ Poffenberger
> what the problem was, but I can't remember where. Since it evidently wasn't
> on here, it must've been in a magazine. Since the only technical magazine
> (only because it's the only computing magazine that I read :) is BYTE, I
The Oct. 31 issue of EE times had a good description of the bug.
Russ Poffenberger DOMAIN: pof...@sj.ate.slb.com
Schlumberger Technologies UUCP: {uunet,decwrl,amdahl}!sjsca4!poffen
1601 Technology Drive CIS: 72401,276
San Jose, Ca. 95110
(408)437-5254
-------------------------
In a dictatorship, people suffer without complaining.
In a democracy, people complain without suffering.