Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!think.com!zaphod.mps.ohio-state.edu!sdd.hp.com!elroy.jpl.nasa.gov!ncar!midway!clout!chinet!saj
From: saj@chinet.chi.il.us (Stephen Jacobs)
Newsgroups: comp.sys.atari.st
Subject: Re: Atari Mortis
Summary: Some particulars
Keywords: history, cheerleading, admonitions
Message-ID: <1991May29.141219.3600@chinet.chi.il.us>
Date: 29 May 91 14:12:19 GMT
References: <1991May19.035413.14005@chinet.chi.il.us> <1111@stewart.UUCP> <1991May28.180143.4644@colorado.edu>
Organization: Chinet - Chicago Public Access UNIX
Lines: 69

In article <1991May28.180143.4644@colorado.edu> chuj@horton.Colorado.EDU (CHU JEFFREY) writes:
>In article <1111@stewart.UUCP> jerry@stewart.UUCP (Jerry Shekhel) writes:
>>saj@chinet.chi.il.us (Stephen Jacobs) writes:
>>>
>>>Sure enough, in selected applications,
>>>the TT beats the pants off anything based on an Intel chip.  In a lot of
>>>other applications, it runs pretty much even with a 25 MHz 80486.  And the
>>>price is in the 80386 range.
>>>
>>
>>Not to flame you or anything, but I seriously doubt all three claims.  Perhaps
>>you could provide some numbers to substantiate them?
>
>I doubt it too, there was a demo of the TT here by an ATARI club, the
>representative said the TT 68030 was equal to a 386-20 machine and only
>2 MIPS, I found this hard to believe since there was so much talk of it,
>I still think it does close to 8MIPS, but since there is so much people
>disagreeing with the performance of the TT, I don't know what the real
>MIPS on it.  Also does anyone know how many MFLOPS the TT does?  Like I
>said the TT is not equivalent to the i486 and probably not the new 386-40.
>The 386-33 is running 7.92 MIPS and the i486-25 is 11.1 MIPS.
>THIS IS NOT A FLAME (I am a ATARIAN and probably always will be)
>
>
>							Jeff

Some particulars.  Dynacadd ($1000 list, but available for about $550) handles
drawings as well as Auto CAD.  On a TT with a math coprocessor, it does an
arbitrary rotation of the semi-standard space shuttle drawing in essentially
human reaction time (not quite, but too fast to time conveniently).  Last I
heard, a '486 running Auto CAD took about 2 seconds.  Pagestream and Calamus
run at practical speeds on the TT; they are highly usable (I use Pagestream on
a Mega, myself, and while it's highly functional, some text operations are 
slower than I'd like).  The comparable programs for Intel chips, whatever their
merits, just don't seem to get used.  Similarly, because of the Atari-MIDI
connection, there are good heavy-duty MIDI programs for the TT; in the 
'mainstream', the word seems to be that for MIDI you get a Mac.

Pricing: we just saw a quote of $2100 for a stripped TT.  That's '386 pricing.
Similarly, the full configuration prices for the TT seem to be in the low
$3000 range.  Same comment.  

Routine applications: my big routine application is chromatographic data 
processing.  Big disk files going through once.  Anything faster than a '386
25 runs i/o bound on this one, so all fast machines look the same.  Any big
math operation that has an inner loop that takes up residence in on-chip cache
(and that's a lot of the market for cycles right there) is going to turn over
amazing numbers of MIPS on any fast processor.  The relative amazingness is
going to depend on machine cycles per instruction (slight advantage to the '486
I believe) and efficiency of data addressing (slight advantage to the '030).

About that 2 MIPS for a 32 MHz '030 box: the TT needs some configuring to work
at its best.  This is new for Atari machines, but is old hat for other micros
(and minis).  Run a program with bad locality of addressing entirely out of
RAM shared with video, maybe you can make it run that slowly.  Unfortunately,
unless someone tells the system not to, that's the RAM the program will use.
And locality just wasn't the issue in 68000 programming that it is in 68030
programming, so lots of existing programs aren't ideal for the TT.  Which comes
to the old lament that the only real benchmark is YOUR application.

The BIG problem with the TT (and all machines Atari) is that the VARIETY of
applications just isn't out there.  The big applications are covered, but not
with a whole lot of different choices.  And Atari doesn't make it easy to 
develop for their newer machines (although they try to make up for it with 
their help once an application is developed).

I hate to argue in public.  I'd have done this in email, but I think maybe this
discussion might be worth keeping going.
                                            Steve