Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!apple!autodesk!peb
From: peb@Autodesk.COM (Paul Baclaski)
Newsgroups: comp.sys.next
Subject: Re: Next, 40, Sparc, 2
Summary: cool your jets
Message-ID: <779@autodesk.COM>
Date: 21 Nov 90 21:29:37 GMT
References: <11090@pt.cs.cmu.edu> <1027@toaster.SFSU.EDU>
Sender: news@Autodesk.COM
Lines: 59

In article <1027@toaster.SFSU.EDU>, eps@toaster.SFSU.EDU (Eric P. Scott) writes:
> In article <30017@boulder.Colorado.EDU> fozzard@alumni.colorado.edu
> >This statement would be better applied to the aging 68xxx CISC architecture,
> >not the deliberately *scalable* RISC SPARC design.
> Religious dogma.  Ho-hum.

Perhaps you don't understand what scalable means.  Sun is committed to 
doubling performance for a constant price every 2 years and they are
doing it.  At least 30% of their customers are software developers
who really like to get faster compiles and higher MIPS does a whole
lot in this area.  CISC machines are not scalable so they take 
considerable effort to crank up the clock speed.

The old rule of having a megabyte per MIP is a pretty good one.  I 
just got 8 more meg in my sparcstation at work and now the disk isn't
busy all the time.  Now that I have 16 meg, I will be checking out
OpenWindows; I think it really requires that much memory.

> >that can take days - don't tell anyone around here that 28 MIPS isn't
> >meaningful.
> And yes, I've seen NextStep run on IBM RS/6000s.  RISC machines.
> RISC machines that leave your Suns in the dust.  So there.

You are correct that the IBM PowerStations (where did they get that name?)
are faster (and more expensive), especially in floating point.  Their
machines are superscalers which execute 1.2 instructions per cycle, so
they can squeeze a little more out of the clock cycle.  Now that Sun
has customers interested in floating point, they are designing machines
that do that--4MFLOPS for the sparcstation 2.  Note also that for
random disk access tests, the sparcstation 1 (as reviewed in Unix Review),
had better performance by an order of magnitude when compared to systems
designed to be servers (a machine from MIPS).

Although there are lots of people who seem to think MIPS are great
comparison factors, people might not be a dumb as you think--for 
instance, if you already use a vendors workstation and they make it
twice as fast in MIPS, then the new machine will be faster for you.
When comparing machines from different vendors, lots of other things
come into play.  For instance, I'm trying to decide whether I should
get a NeXT or a Sparcstation for home use.  If I get a Sun, I can 
get all the benefits of a world class operating system and I know
where they are going.  If I go with NeXT, then I get NextStep which
is better than Guide (Sun's tool for creating user interfaces) because
it is easier to modify old programs with Guide and NextStep also has
some neat button arrangment methods.  It seems that both machines have
the problem that software vendors charge a lot for software (based
on what you paid for the computer, not based on what it costs to 
develope the software...).  NeXT still has a big hole--no quality
debugging environment.  It is amazing that they have not done anything
in this area in 2.5 years.

NeXT will probably always lag behind Sun in MIPS, but their 
philosophy of having a better common denomenator (what comes 
bundled, hardware and software wise) is probably their most 
attractive quality.

Anyway, enough verbiage on this dead horse.

Paul