Re: Apple and the Holy War

[pausch@saafNOSPAM.se (Paul Schlyter)]

In article <391C0FDF.E58FC8D9@inetnebr.com>,
Roy and/or Janet Miller  <millers@inetnebr.com> wrote:
 
> Supertimer wrote:
> 
>>  But the fact of
>> the matter is the majority of PC users did not plug a
>> math coprocessor into the socket even though their
>> PC had one.
> 
> On true blue IBM PCs, you not only had to add a mathco, you had to replace
> the 8088, because the one that came did not work with the 8087 - at least
> that's what the IBM math co upgrade pages in the IBM Guide to Operations
> manual say
 
True!  And this was valid not just for the IBM PC but for the clones
as well, even though it was rarely explicitly mentioned.  The story
here is quite interesting: the presence of the 8087 imposed somewhat
more severe restrictions on timing/etc on the 8088 chip.  Many 8088
chips failed to meet these requirements, but since most people didn't
use the 8087, and since these 8088's worked fine with no 8087
present, they were used anyway rather than discarded.  To be on the
safe side, IBM supplied an "8087 certified" 8088 along with their
8087's.
 
However, not all of those 8088's running alone worked bad with the
8087.  When I added an 8087 to my XT clone, I didn't replace the
8088.  It worked anyway, so I guess I was lucky.  Later I tried to
replace the 8088 with a NEC V20 chip, and the V20 worked only if the
8087 was absent.  At that time NEC also promised an 8087 replacement
- I waited for it awhile, but I never saw it and eventually gave up
waiting for it.
 
Sometimes I swapped chips in my XT: I usually had the 8088+8087
inserted, but sometimes I removed them and inserted the NEC V20
instead.  Not because the marginal 10-15% speedup for CPU intensive
code which the V20 yielded, but because I was interested in another
feature of the V20: it could emulate the 8080 as well!  Yes, the 8080
is Intel's old 8-bit CPU for which the CP/M OS was written.  There
were CP/M emulators available for the PC quite early: some of them
used the native 8080 mode of NEC V20 if they detected the presence of
the V20.  When I ran such an emulator, CP/M programs ran some 20
times faster with the NEC V20 running in native 8080 mode than with
software 8080 emulation on the 8088.  Later I got an 80386 (with an
80387 of course!) -- then the 8088+8087 got permanently removed from
my XT clone and replaced with the NEC V20 - and that NEC V20 still ran
CP/M programs faster than software 8080 emulation on the 80386!
Some years later still, software 8088 emulation on the 80486
finally did beat the 8080 native mode of the NEC V20 in performance.
 
The NEC V20 used the quite useless "POP CS" instruction of the
8088 to enter 8080 native mode.  Later, Intel removed the "POP CS"
instruction from the instruction set (I believe this happened with
the 80286).
 
-- 
----------------------------------------------------------------
Paul Schlyter,  Swedish Amateur Astronomer's Society (SAAF)
Grev Turegatan 40,  S-114 38 Stockholm,  SWEDEN
e-mail:  pausch at saaf dot se   or    paul.schlyter at ausys dot se
WWW:     http://hotel04.ausys.se/pausch    http://welcome.to/pausch