rob@lafayet.UUCP (Rob Freyder) (08/13/90)
Greetings Folks, I have about 14 HP Vectra 386/20 systems all with 387 co-processors. We run SCO Xenix currently and have had a recurring problem that up until lately has baffled us.... Whenever we would run SCO Professional (spreadsheet) accessing the tape drive or floppy disks with tar or dd (the only ones i tested) the system would "hang" without a single message. I consulted the SCO database and found that I was not alone in this dilemma. There have been many others that have had the same problem... I am not quoting but it goes something like this. If you are in protected mode and you are executing a 32 bit fp instr and access a DMA device the CPU will "freeze" or "hang up". This is only happens with certain releases of the chip and is apparently not a problem with newest release. SCO has a workaround that works on some machines (of course not on ours) ... IF it doesnt work on yours then you can disable the 387 until you are able to replace the CPU... 8-( Intel has documented this and I believe its referred to as "Erratum 21" I need pointers on identifying these "defective" 386 chips and was wondering if anyone has leads on diagnostic software that would verify my suspicions regarding this "hanging" of the system... Also .. If I make a visual inspection of the chips what markings would tell me that it was the type that "hangs" in protected mode. And finally, ...What should we replace these with ? Cost ? Has anybody else out there been through this ?? Help ! BTW , this problem doesnt manifest itself under DOS because, of course, DOS doesnt run in protected mode. Thanks in advance for any pointers, replies or suggestions that you might have. Rob -- Rob Freyder Core Laboratories a division of ____ ____ ____ Western Atlas International a \ \ / /\ / /\ Litton/Dresser Company \ \/ / \ / / \ (318) 235-9431 rob@lafayet.uucp \ / / \ / /\ \ -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= \/___/ \/___/ \___\ ...!uunet!rouge!lafayet!rob
shef@ipla01.hac.com (Ian B. Shef) (08/14/90)
In article <504@lafayet.UUCP> rob@lafayet.UUCP (Rob Freyder) writes: > > [ Many lines describing a Unix hang problem deleted ] > >Intel has documented this and I believe its referred to as "Erratum 21" > > [ More lines deleted ] > There is a company which frequently advertises in _EDN_ magazine which has a hardware fix for your problem (if indeed it is errata 21). For example, see _EDN_, March 15, 1990, p. 217: " The FIX-80386 solves the Errata 21 problem that is showing up on many PC's. If your PC locks up when running UNIX or memory extenders in MS-DOS you will need this part. The part is placed between the 80386 and its socket... " The company is Ironwood Electronics, P.O. Box 21151, St. Paul, MN 55121. Phone number 612-431-7025. I neither work for Ironwood Electronics, nor have I had occasion to use their products. Hope this helps. -- Ian ------------------------------------------------------------------------------ -- Ian Shef Internet: shef@ipla01.hac.com [128.152.40.86] Hughes Aircraft Co. 268/W62 uucp: ...!usc!hacgate!ipla01!shef (maybe?) 8433 Fallbrook Avenue phone: 818-702-4365 Canoga Park, CA 91304-0445 Any opinions expressed are author's own.
johnl@esegue.segue.boston.ma.us (John R. Levine) (08/14/90)
Executive summary: you lose.
Long version: Any 386 chip that doesn't say 386DX is subject to this problem.
Any 32 bit paging system, e.g. Xenix, Unix, OS/2 2.x can fail. For a while,
Intel had an informal chip swap programs for system vendors, but they insist
it's too late now since fixed chips have been available for over a year.
Intel has a support phone number for the 387, which answers with a recording
that starts "We're incredibly busy now ...," even at 3:00 AM. Among the
other things the recording tells you is the number of their fax-back machine,
which you call from a touch tone phone, and it turns around and calls your
fax machine back and sends you helpful support documents. One of the
documents (which, of course, is in a file cabinet 300 miles from here) talks
about it and mentioned that someone has a little daughtercard that avoids the
problem. There is a directory of documents that you can retrieve, and the
relevant one is called something like the 386 system compatibility list.
Note that even Intel motherboards have this problem, I have one and they're
not willing to do anything about it.
If you find the source of the daughtercards, or alternatively someone who
will actually sell a single 386/25 chip, I'd love to hear about it.
Regards,
John Levine, johnl@esegue.segue.boston.ma.us, {spdcc|ima|world}!esegue!johnldavidsen@crdos1.crd.ge.COM (Wm E Davidsen Jr) (08/14/90)
In article <9008131759.AA16793@esegue.segue.boston.ma.us> johnl@esegue.segue.boston.ma.us (John R. Levine) writes: | fax machine back and sends you helpful support documents. One of the | documents (which, of course, is in a file cabinet 300 miles from here) talks | about it and mentioned that someone has a little daughtercard that avoids the | problem. Bell Technologies (not part of Intel) was selling a board for that, cost $99 as I recall. Their phone number used to be 1-800-FOR-UNIX. | If you find the source of the daughtercards, or alternatively someone who | will actually sell a single 386/25 chip, I'd love to hear about it. There were a couple of chips advertised in misc.forsale recently. -- bill davidsen (davidsen@crdos1.crd.GE.COM -or- uunet!crdgw1!crdos1!davidsen) "This is your PC. This is your PC on OS/2. Any questions?"