Path: gmdzi!unido!mcsun!uunet!maverick.ksu.ksu.edu!zaphod.mps.ohio-state.edu!
uwm.edu!rutgers!cbmvax!snark!eric
From: e...@snark.uu.net (Eric S. Raymond)
Newsgroups: comp.arch,alt.folklore.computers
Subject: Why IBM went for the 8088 -- The Untold Story
Message-ID: <1WWjqN#0SBCWN5sLFYH9X6tz009zLSK=eric@snark.uu.net>
Date: 22 May 90 20:03:04 GMT
Followup-To: comp.arch
Lines: 66
Posted: Tue May 22 21:03:04 1990
Once upon a time back the early Eighties, I worked for a micro software
house which shall remain nameless (it was a literal basement sweatshop run by
a gonif with a penchant for folding companies when his debts piled up to
high, then restarting business under a new and equally sleazy banner). I
was young and innocent then...cranking out code for TRS-80s and Osbornes
and legions of forgotten CP/M machines. I started there just before the IBM PC
came out.
Anyhow, the company had a secret agent inside IBM -- a rabbity little nerd
we called "Deep Disc" (I met him twice, but never knew his real name) who used
to pass us confidential IBM technical memoranda (I still have some of them).
Goddess knows what he got out of this beyond the Excitement Of It All.
This had a couple of interesting results. One is that I was one of a really
tiny number of people who got a look at MS-DOS before the PC was released,
running on an odd little non-IBM 8088 box the gonif bought for the purpose.
Another is that the gonif was able to get his lunch hooks on three of the
first six IBM PCs to hit the Eastern Seaboard. One of them (PC #660-something)
became my machine, and my girfriend uses the keyboard from it to this day
(the rest of the machine having been replaced by clone parts over the years).
And, finally, it means that we got the inside scoop on why IBM went with the
8088. It seems that IBM chartered two design groups to build them a cheap
personal micro aimed at eating the Apple II's lunch. One group designed a
machine around the MC6800; the other used the 8088, apparently to go with
an assumption that the OS would be a mechanical port of the Z-80's CP/M.
These concepts were known inside IBM as the "high-ball" and "low-ball" designs.
(The CP/M deal, legend has it, went sour later in the process because Gary
Kildall decided to enjoy the perfect flying weather the day the IBM people
had plans to meet with him and blew them off. Eventually, there was a CP/M
for the IBM-PC; I ran it once or twice. It sank without trace.)
Well, we all know (and curse the fact that) the 8088 version was chosen. What
everybody gets wrong is why. It had nothing to do with the relative cost of
the candidate microprocessors or availability of support chips (though that's
the myth IBM and Intel later encouraged). It wasn't the result of any high-
level collusion between Intel and IBM, nor of ineptness at Motorola nor any
murky personal jim-jam among the individual decision-makers involved.
No -- according to Deep Disk (who I believe was placed to know) the reason
was much simpler and much more in line with IBM's strategic traditions going
back to the days of System/360. They picked the low-ball 8088 design to avoid
cannibalizing sales of the next machine further up.
Unfortunately for us all, the "next machine farther up" was a doomed turkey
called the System/23, a dedicated word-processing machine like the old Lanier
boxes that ran a customized version of Z-80 CP/M off of *slow* 8-inch floppies.
IBM chose the 8088 design to be *less capable* than the System/23.
It's hard to remember now just how stupid the original PC-1 machines were,
because IBM doubled the original floppy capacity and speed and introduced the
64K motherboard quite soon after they first came out, after it became clear
that the System/23 was a goner (for reasons that I think had less to do with
the PC than with the System/23's own high price and losing traits). But they
were really, truly wretched -- designed that way on purpose as a compromise
between elements in IBM that saw a chance to eat the Apple II's lunch with a
faster VisiCalc box, and others who believed the PC would "never amount to
anything" and were concerned primarily with protecting the System/23.
So there you have it, fans. IBM chose the 8088 to avoid competing with a
wretched failure of its own make -- one that was later so thoroughly forgotten
that the true story of the PC's origins was itself obscured. But this is how
it happened according to someone who was there.
--
Eric S. Raymond = ...!uunet!snark!eric (mad mastermind of TMN-Netnews)
Path: gmdzi!unido!mcsun!uunet!mailrus!news-server.csri.toronto.edu!utgpu!utzoo!
sq!ian
From: i...@sq.sq.com (Ian F. Darwin)
Newsgroups: comp.arch,alt.folklore.computers
Subject: Re: Why IBM went for the 8088 -- The Untold Story
Summary: Don't neglect Seattle Computer or the S100 bus.
Keywords: S100 IEEE-696 SC-DOS Seattle MS-DOS IBM_PC
Message-ID: <1990May23.154706.16476@sq.sq.com>
Date: 23 May 90 15:47:06 GMT
References: <1WWjqN#0SBCWN5sLFYH9X6tz009zLSK=eric@snark.uu.net>
Organization: SoftQuad Inc., Toronto Canada
Lines: 44
Posted: Wed May 23 16:47:06 1990
Eric S. Raymond (e...@snark.uu.net) writes a very interesting tale
of the origin of the IBM_PC's segmented braindeath. I can only add one
minor quibble, by way of reminding people of another aspect that
is often overlooked:
> ... One is that I was one of a really tiny number of people who got a look
> at MS-DOS before the [IBM]PC was released,
> running on an odd little non-IBM 8088 box the gonif bought for the purpose.
In fact, dozens or hundreds of people used SC-DOS before IBM bought it from
Microsoft who in turn bought it from a small company named Seattle
Computing. Microsoft changed the name to obliterate the history, and
(somewhat later) stole the notion and notation (verbatim but for one
typographical error, the \ as a directory delimiter) for tree-structured
directories. SC-DOS was originally a clone of CP/M, but done for the 8086,
the 8088's faster sibling, that was becoming the CPU base of many of the
forgotten micro makers that IBM drove out of business in the first few years
of selling PC's.
Many of those micro makers based their products on what was then called the
S100 bus. Early S100 systems were plagued by inter-vendor incompatibilities,
so an IEEE standards body cleaned up the standard, and vendors were (a very
few still are, I believe) producing IEEE-standard boards which did in fact
interoperate. But by this time the IBM_PC bus had too much momentum for it to
matter. What is interesting is that the IEEE standard included a definition
for "DMA" (direct memory access) that was in turn based on earlier
IBM work, *and* was practically a blueprint for IBM's MCA.
So, the pre-IBM_PC S100 bus contributed large parts of both the software for
the IBM_PC and the hardware design for IBM's current offerings in that
field.
Ian Darwin
i...@sq.com
References: leaf through back issues of \fIMicrosystems\fP from 1982-1983
for the Seattle Computer stuff. My copies of this, and the book by Sol Libes
& Mark Garetz on Interfacing the S100 Bus (approximate title), are both at
home buried under several layers of boxes while the house is renovated, so I
can't give specific issues or page numbers. If you need citations, go to the
library, or ask me again in about three months.
--
#exclude <stddisclaim.h>
Path: gmdzi!unido!mcsun!uunet!cs.utexas.edu!rutgers!cbmvax!snark!eric
From: e...@snark.uu.net (Eric S. Raymond)
Newsgroups: comp.arch
Subject: Re: Why IBM went for the 8088 -- The Untold Story
Message-ID: <1WXlwg#5vmkWO5PBg8W389WlZ58SnBR=eric@snark.uu.net>
Date: 24 May 90 15:02:45 GMT
References: <1WWjqN#0SBCWN5sLFYH9X6tz009zLSK=eric@snark.uu.net>
<11733@cbmvax.commodore.com>
Lines: 11
Posted: Thu May 24 16:02:45 1990
In article <1WWjqN#0SBCWN5sLFYH9X6tz009zLSK=e...@snark.uu.net> e...@snark.uu.net
(Eric S. Raymond) writes:
>And, finally, it means that we got the inside scoop on why IBM went with the
>8088. It seems that IBM chartered two design groups to build them a cheap
>personal micro aimed at eating the Apple II's lunch. One group designed a
>machine around the MC6800; the other used the 8088, apparently to go with
>an assumption that the OS would be a mechanical port of the Z-80's CP/M.
>These concepts were known inside IBM as the "high-ball" and "low-ball" designs.
Oops. Typo alert. Make that `MC68000'.
--
Eric S. Raymond = ...!uunet!snark!eric (mad mastermind of TMN-Netnews)
Path: gmdzi!unido!mcsun!uunet!aplcen!samsung!cs.utexas.edu!execu!sequoia!attdso!
hriso!atti07!ulysses!dptg!pegasus!dmt
From: d...@pegasus.ATT.COM (Dave Tutelman)
Newsgroups: comp.arch,alt.folklore.computers
Subject: Re: Why IBM went for the 8088 -- The Untold Story
Keywords: S100 IEEE-696 SC-DOS Seattle MS-DOS IBM_PC
Message-ID: <4790@pegasus.ATT.COM>
Date: 25 May 90 11:59:44 GMT
References: <1WWjqN#0SBCWN5sLFYH9X6tz009zLSK=eric@snark.uu.net>
<1990May23.154706.16476@sq.sq.com>
Reply-To: d...@pegasus1.ATT.COM (Dave Tutelman)
Organization: AT&T Bell Labs - Lincroft, NJ
Lines: 52
Posted: Fri May 25 12:59:44 1990
In article <1990May23.154706.16...@sq.sq.com> i...@sq.sq.com (Ian F. Darwin) writes:
>Eric S. Raymond (e...@snark.uu.net) writes a very interesting tale...
>I can only add one minor quibble...
... and I'd like to add a minor quibble or two. In general,
Ian's article was on the money.
>In fact, dozens or hundreds of people used SC-DOS before IBM bought it from
>Microsoft who in turn bought it from a small company named Seattle
>Computing. Microsoft changed the name to obliterate the history...
I'm un-fond of Microsoft myself, but let's be fair about
motive. Microsoft bought the rights (and the code) from
Seattle Computing (SC...), and the resulting product was
Microsoft's (MS...). I don't think "obliterating history"
was ever a consideration in the name change.
>SC-DOS was originally a clone of CP/M, but done for the 8086,
>the 8088's faster sibling...
You're being slightly misleading about the 8086 vs. 8088 here.
They're software-identical, and both chips were available in
a variety of clock rates. At the same clock rate, the only
difference is that the 8086 has a 16-bit bus to the 8088's
8-bit bus. Your impression probably results from the facts that:
- Many 8086 machines of the era were made with the
8 MHz part, whereas IBM chose either cheap chips
or very conservative design to run their 8088 at
4.7 MHz.
- The difference in bus width alone makes 8086 machines
run "typical" programs about 40% faster than 8088
machines at the same clock rate.
>...that was becoming the CPU base of many of the
>forgotten micro makers that IBM drove out of business in the first few years
>of selling PC's.
In fact, there were some 8086-based clones, whose market was
made, not killed, by the IBM PC. Consider the AT&T PC6300 (actually
an Olivetti machine). It used an 8 MHz 8086, and was as
faithful a clone as most of the clones of that generation.
It not only ran the software for the IBM, but accepted its
(8-bit bus) add-on boards. Yes, there were a few incompatibilities,
but there were similar "bugs" in other first-try clones using
the 8088.
Does anyone know of any other 8086-based XT clones (not made by
Olivetti), or is this the only counter-example to Ian's
statement?
+---------------------------------------------------------------+
| Dave Tutelman |
| Physical - AT&T Bell Labs - Lincroft, NJ |
| Logical - ...att!pegasus!dmt |
| Audible - (201) 576 2194 |
+---------------------------------------------------------------+