Will@cup.portal.com (Will E Estes) (10/11/89)
A vendor just solved a compatibility problem I was having between its motherboard and several ESDI controllers by replacing the '386 CPU on the board with an *older* version of the chip; I was wondering if someone here could give me information on the two versions of the chip and tell me what the differences are. The original chip in my system was labeled as follows: 386DX-25 IV SX218 L9330799 Copyright (C) '85 The older '386 chip which was used as a replacement in my system was labeled as follows: 386DX-25 L9021187 Copyright (C) '85 '86 The claim made by the vendor's technician was that the version IV chip that had been used on the motherboard I purchased has a known bug with the pipeline instruction. Apparently the older version of the chip does not? In any case, can someone explain to me what is the difference in capability/known bugs between the two chips I describe above? I'm very curious as to why an ESDI controller would not work with a later version of the Intel chip (I would have guessed that the '386 was a more mature product by now). Thanks, Will
plim@hpsgpa.HP.COM (Peter Lim) (10/12/89)
> The original chip in my system was labeled as follows: > > 386DX-25 IV > SX218 > L9330799 > Copyright (C) '85 > > The older '386 chip which was used as a replacement in my system > was labeled as follows: > > 386DX-25 > L9021187 > Copyright (C) '85 '86 > Older ???? From the Copyright date you wrote up here, the replacement chip seems to be newer. The only problem I know with older (not newer) 80386 chips was that some 32 multiply instruction doesn't work properly. That's about all I can say ........ Regards, Peter Lim. HP Singapore IC Design Center. E-mail address: plim@hpsgwg Snail Mail address: Peter Lim Hewlett Packard Singapore, (ICDS, ICS) 1150, Depot Road, Singapore 0410. Telephone: (065)-279-2289
Will@cup.portal.com (Will E Estes) (10/14/89)
Well, here's the scoop on the bug with the '386 chip from Intel. Still
doesn't explain my problem, but it does give more information on the
problem in general:
Date: 13-Oct-89 09:46 PDT
From: >CSI:@INTEL
Subj: 386 bug
To: Will Estes
Gary Brown
Joe Beauchamp
Russ Niklos
Carl Rock
From: Bob Albers @INTEL
Subj: 386 D-0 step CHMOS-IV pipelined bus mode bug
During what should have been a routine shrink of the 386DX from
CHMOS-III to CHMOS-IV a new bug was introduced that affects the ability
of the 386 to operate correctly in pipelined bus mode. The bug only exists
when the on-chip paging unit is enabled AND address pipelining is enabled on
the external bus AND some other exotic conditions are also true. Most users
Enter command or <CR> for more!
EasyPlex
will not see the problem because most DOS software does not use paging; also
most higher speed 386 designs don't enable pipelining because they have an
external cache memory which makes pipelining unnecessary and more difficult
to implement. All 386 board vendors were notified about this erratum by
December, 1988 at the latest and had the option to disable pipelining
on their boards or to buy the CHMOS-III version. Please refer to the list
below to determine the version of the 386 that you have in your system.
Definitions: CHMOS-III 1.5 micron CMOS process
CHMOS-IV 1.0 micron CMOS process
D-0 step Version on CHMOS-III process
does NOT have pipeline bug.
Version on CHMOS-IV process
has pipeline bug.
D-1 step Version on CHMOS-IV process
that fixes pipeline bug.
386DX-25 IV CHMOS-IV D-0 step
SX133 <--has pipeline bug
386DX-25 CHMOS-III D-0 step
SX215 <--some older yet identical units
<--were without an S#
<--never had pipeline bug
386DX-25 IV CHMOS-IV D-1 step
SX218 <--pipeline bug fixed
Since the message posted by Will Estes indicated that his board
was changed from an SX218 to an SX215 CHMOS-III part I'm confused. There
are no known errata or spec differences between these two versions.
The CHMOS-IV parts are generally faster though and this could have
caused a marginal board design to become flakey with one version or
another.