Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!ucbvax!bloom-beacon!apple!rutgers!bellcore!texbell!killer!elg
From: elg@killer.Dallas.TX.US (Eric Green)
Newsgroups: comp.arch
Subject: Re: 80486 vs. 68040 code size
Message-ID: <8081@killer.Dallas.TX.US>
Date: 12 May 89 01:13:10 GMT
References: <922@aber-cs.UUCP>
Distribution: eunet,world
Organization: The Unix(R) Connection, Dallas, Texas
Lines: 62

in article <922@aber-cs.UUCP>, pcg@aber-cs.UUCP (Piercarlo Grandi) says:
> Also, the 68020 was introduced not much before the 386, and certainly the
> 386 has been around far longer than the 68030. Or I am goofing?

You're goofing. 68020's have been around at least since 1986 (which is
when I saw my first one, a CCA add-on processor for the Amiga). And
Suns 3s had them before then.  The '020 became available a bit over a
year after the '286, I seem to recall.  '386 machines only recently
have become widely available, i.e., the current availability of the
'386 is about the same as the availability of the '020 in 1986.


The '030, on the other hand, came out slightly after the '386. It's
only now coming into widespread use, about a year after the '386 did.

As for the price difference: Apollo's low-end '030 machine is
advertised for around $6k, which is about what a competitive '386
machine (4 megs of RAM, super-VGA, etc.) would cost. Remember, these
'030 machines you're seeing for sale are Unix workstations, with
multimegabytes of RAM, and a high-speed network card or hard drive,
running Unix, with large (expensive) displays. These aren't the "price
buster" CGA 1 megabyte MS-DOS '386 machines you can get for $3000 out
of Computer Shopper.... to compare, you have to add a bit (and you
still won't get there, because '386 Unixes generally don't do
full-blown windowing and networking). It's interesting to note that
Apollo's low-end '030 machine costs about the same as Apple's '020
machine (is this a case of comparing Apples and, uh, Apollos? ;-).


> on the few 68030 machines around (Next, Macs, Suns), it seems that one
> statement I have read (68030 == 68020+10/15%) is quite reasonable

Figures I've seen indicate about a 20-30% improvement. Note that
performance gains can be sapped by high-overhead operating systems
(e.g. Mach on the NeXT) or by anemic memory systems (e.g. Mac IIx).
You've benchmarked 68030-based Suns??? Considering that they only
recently introduced them, you've been working fast! (the Sun's got
about the same overhead as a NeXT anyhow, except it doesn't have the
object-oriented layer to the user interface). 
    A fair comparison would be, hmm, a generic '386 clone with AT&T's
Sys V and Unix compiler against a generic '020/'030 VME Unix system
running generic AT&T Sys V. I've never seen such a comparison done.
All I've ever seen is comparisons between apples and oranges, uh, Sys
V/Microsoft C vs. BSD4.2/PCC. Each machine would have to have the same
amount of main memory, the same amount of cache, and the same clock
speed. Then we could decide.

> This would put the 68030 on a par with the 386 in speed. As to the

It would, sort of -- I believe they both do a fetch in 2 cycles?
EXCEPT -- 33mhz 68030s are available NOW. For the '386, it's "Real
Soon Now". In general, Motorola has been ramping the clock up on the
'030 faster than Intel has on the '386. Of course there's a limit to
this -- how fast can your cache be accessed? But by that time, the
'040 will be out, hopefully with increased bus bandwidth e.g. Harvard
memory architecture.

--
|    // Eric Lee Green              P.O. Box 92191, Lafayette, LA 70509     |
|   //  ..!{ames,decwrl,mit-eddie,osu-cis}!killer!elg     (318)989-9849     |
|  //    Join the Church of HAL, and worship at the altar of all computers  |
|\X/   with three-letter names (e.g. IBM and DEC). White lab coats optional.|