🚀 go-pugleaf

In article <120@utastro.UUCP>, jk@utastro.UUCP (John Krist) writes:
> I would like to see some talk about 32032 vs. 68020 vs. 80386 here in
> .16k land.  The people at Definicon say that the 68020 still has bugs
> and that Motorola's math processor doesn`t work well.  It seems that
> the 32032 is the most VAX-like.  Would this make it better for those
> surrounded by VAXen?

If you want the full story you should also read net.micro.68k and
net.arch, where these questions are being discussed (sometimes to
death).  The 68020 in the Sun-3 I'm typing on now has a few minor bugs,
but none that the compilers have to know about or that affect user
programs.  Last I heard, the National compilers have lots of hacks to avoid
instructions and sequences that tickle chip bugs.  (The worst of it is,
Motorola supplies a bug list to its customers, under nondisclosure;
National won't, and even took action against some of its own employees
who were telling customers about chip bugs.)

The 68881 math processor works fine and does nice 80-bit math in full
IEEE glory; the 16081 only does 64-bit math, and is much slower.

The 80386 looks like a serious contender to the 68020, though nobody I've
ever talked to, including people working on porting things to it, has
seen one.  (They use software simulators.)  When it is released, and
the operating system mods are done (not in the first release, I bet),
it will be able to run MS-DOS programs and such at the same time as
Unix -- which is a big win for many people.  On the other hand, I
bought the Sun-3 on the open market and got it a month ago and it runs
4.2BSD much faster than a Vax 780.  How long will it take to get a 4.3
port for the 386?

The 32032 is dead in this market; it's only 30% faster than a 16032,
which puts it right in the 68000 ballpark -- if you can get a bug-free
chip that runs at its rated clock speed.  (Haven't talked to anybody
who has seen one of those either.)  The 32332, well, it's faster than a
32032, but slower than both 68020 and 80386, so only people locked-in
to National chips will buy it.

I'd say the best micro for people who really like Vaxen is the microvax-2.


>        [DSI-32] WHETSTONE benchmark:
>            C: optimized - 4.11 sec
>               non-opt   - 4.40 sec
>          F77: optimized - 4.45 sec
>               non-opt   - 4.45 sec

Hmm, that's quite an optimizer.

[]

	"The worst of it is, Motorola supplies a bug list to its
	customers, under nondisclosure; National won't, and even took
	action against some of its own employees who were telling
	customers about chip bugs."

I don't know how this rumor got started, or why people keep repeating
it.  National distributes regular chip bug reports; I've been getting
one every few months ever since the 16032 began sampling.  We've also
discovered and contributed some novel bugs, and they've all shown up in
the reports, so I don't think they're holding anything back.

  -=- Andrew Klossner   (decvax!tektronix!tekecs!andrew)       [UUCP]
                        (tekecs!andrew.tektronix@csnet-relay)  [ARPA]

In article <314@l5.uucp> gnu@l5.uucp (John Gilmore) writes:
>...  Last I heard, the National compilers have lots of hacks to avoid
>instructions and sequences that tickle chip bugs.  (The worst of it is,
>Motorola supplies a bug list to its customers, under nondisclosure;
>National won't, and even took action against some of its own employees
>who were telling customers about chip bugs.)

Maybe that's why the current bug list was published in Byte...  Current
32032's (rev N) do not need 'lots of hacks' -- 'last I heard', in your case,
must have been when they were producing rev K chips, at which time 68020's
were not in general supply.  And Tek, as a National customer, has complete
access to Nati bug lists.  Who have you been talking to?  Better get a new
source!

>The 32032 is dead in this market; it's only 30% faster than a 16032,
>which puts it right in the 68000 ballpark -- if you can get a bug-free
>chip that runs at its rated clock speed.  (Haven't talked to anybody
>who has seen one of those either.)

I guess you don't talk too loud, except when posting rumors you can't back up!
There are several around here -- they push about 1250 dhrystones, which is
faster than any 68000.

I won't suggest the 32032 is going to take the world by storm.  The 68020 is
obviously faster.  (Doesn't it also cost about 3-4 times what the 32032 does?)
But John was indulging in exaggeration and rumor that begged rebuttal.

-- 
:::::: Artificial   Intelligence   Machines   ---   Smalltalk   Project ::::::
:::::: Jan Steinman		Box 1000, MS 60-405	(w)503/685-2956 ::::::
:::::: tektronix!tekecs!jans	Wilsonville, OR 97070	(h)503/657-7703 ::::::

In article <314@l5.uucp> gnu@l5.uucp (John Gilmore) writes:
>Motorola supplies a bug list to its customers, under nondisclosure;
>National won't, and even took action against some of its own employees
>who were telling customers about chip bugs.)

We have always had bug lists for the National chips.

>The 68881 math processor works fine and does nice 80-bit math in full
>IEEE glory; the 16081 only does 64-bit math, and is much slower.

How much is much?  May we see your benchmark numbers please?

>The 32032 is dead in this market; it's only 30% faster than a 16032,
>which puts it right in the 68000 ballpark -- if you can get a bug-free
>chip that runs at its rated clock speed.  (Haven't talked to anybody
>who has seen one of those either.)  The 32332, well, it's faster than a
>32032, but slower than both 68020 and 80386, so only people locked-in
>to National chips will buy it.

My number is 503-685-2763.  You can give me a call and talk to me.
Then you will have talked to someone who has seen a 32032 running Unix
and running it at 10Mhz without any kludgey work arounds in the complier.
Again, may we please see your benchmarks between the 32332 and the 68020?

Perhaps your article should have been posted to net.religion.68k.

Cory Barker

In article <978@mako.UUCP> jans@mako.UUCP (Jan Steinman) writes:
>>The 32032 is dead in this market; it's only 30% faster than a 16032,
>>which puts it right in the 68000 ballpark ...
>
>There are several around here -- they push about 1250 dhrystones, which is
>faster than any 68000.

Looks like the 32032 is about the same as a 68010 to me:

grep 68010 dry.c 
 * IRIS-1400	68010-10Mhz	Unix System V	cc		 909	1000
 * ATT PC7300	68010-10Mhz	UNIX 5.2	cc		1041	1111
 * Sun2/120	68010-10Mhz	Sun 4.2BSD	cc		1136	1219
 * MASSCOMP 500	68010-10MHz	RTU V3.0	cc (V3.2)	1156	1238
 * Cyb DataMate	68010-12.5Mhz	Uniplus 5.0	Unisoft cc	1162	1250
 * Sun2/120	68010-10Mhz	Standalone	cc		1219	1315
 * Cyb DataMate	68010-12.5Mhz	Uniplus 5.0	Unisoft cc	1470	1562 S

===========================================================================
Fred Fish    UniSoft Systems Inc, 739 Allston Way, Berkeley, CA  94710  USA
{ucbvax,dual}!unisoft!fnf	(415) 644 1230 		TWX 11 910 366-2145
===========================================================================