Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!uwm.edu!wuarchive!zaphod.mps.ohio-state.edu!uakari.primate.wisc.edu!aplcen!haven!uvaarpa!mcnc!ecsgate!ecsvax!urjlew
From: urjlew@uncecs.edu (Rostyk Lewyckyj)
Newsgroups: comp.sys.amiga
Subject: Re: Real System Comparisons
Summary: Yes indeedee, more of the same!
Message-ID: <1990Aug10.005911.29763@uncecs.edu>
Date: 10 Aug 90 00:59:11 GMT
References: <13466@cbmvax.commodore.com> <13678@cbmvax.commodore.com> <13723@cbmvax.commodore.com>
Distribution: usa
Organization: UNC Educational Computing Service
Lines: 92

It is interesting to note what sections of my message you chose
to comment on and which to ignore. The omitted beginning part
stresses that verifiable benchmark figures of available
machine configurations on floating point numerical problems of a
real size have not been provided by either side of the MSBYS war.
The PC clone side has been perhaps a little more forthcoming than
the Amiga/680xx proponents. Actually on the Amiga side I feel an
active avoidance of providing such results, as evidenced by more
of the same below.
   
In article <13723@cbmvax.commodore.com>, daveh@cbmvax.commodore.com (Dave Haynie) writes:
> In article <1990Aug9.000233.8928@uncecs.edu> urjlew@uncecs.edu (Rostyk Lewyckyj) writes:
> >I have suggested that Spec mark numbers would be very nice
> >I have suggested the following problems: Linpack 300 x 300, 4096 point
> >fft, and a few other such problems.
> 
> An A3000 SPECmark set would be very nice.  However, porting the SPEC benchmark
> suite to the A3000 would be something of a major undertaking.  I don't know of
> anyone here who has that kind of free time, and Commodore isn't currently a
> member of the SPEC group.  I would also like to see someone run the AIM
> workstation benchmark series on the A3000, since these stress multitasking
> performance more than other benchmarks.  But again, other than to acquire
> bragging rights, that's a great deal of work that amounts to a great waste
> of time.  The only people who can justify the work involved would be the
> Marketing group, since they are the only ones who can really get any true
> benefit from such magic numbers.
> 
> And again, you're skipping over 3/4ths of the system when you ask for something
> like Linpack.  I'm sure we could get someone around here to compile and run a
> Linpak, maybe someone already has.  But that ignores many of the things that 
> most PC designers tend to ignore, specifically hard disk throughput, video
> display buffer speed, and expansion bus speed.  Since that last one is my area,
> I have mentioned here both theoretical and actual transfer rates on the Zorro
> III bus.  I have yet to read anything but theoretical figures on MCA or EISA.
> So apparently a good number of people have been a big negligent in their
> benchmarking.  
> 
> >You have asked for the street price of a hypothetical A3000 equivalent
> >machine configuration. Can you supply us with some real performance
> >figures that support your basis for claiming such equivalence.
> 
> We've had plenty of postings by various folks of simple performance benchmarks:
> Dhrystone 2.1, DiskSpeed, MFlops, Sieve, etc.  A3000s run at least as fast on
> these things as other 25MHz 68030 computers, and 80386 machines tend to run
> similar results.  If anything, these kind of tests bias things in the direction
> of 80386 machines, since most have at least 16K of external cache.  Virtually
> all of the simple benchmarks will fit in 16K of external cache, large array
> manipulation tests being the exception.
> 
> >Mr Joon Song answered your posting with his equivalent to your proposed
> >system and a price from Microtimes. Would that system in fact satisfy
> >your criteria?  
> 
> His system missed on two major points: EDSI vs. SCSI disk, and ISA vs. EISA
> expansion bus.  I'm not in a position to judge the quality of a maker I've
> never heard of, and I'm not asking for that 
     
The paragraphs above have been included in the posting for illustration
of the avoidance behaviour involved.
  =-=-=-=-=-=-  This next sentence is worth some thought  -=-=-=-=-=
   
>-- I'm simply looking for a
> valid comparison based only on technical merit.  Caveat Emptor would certainly
   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^  ????? :-)
> apply to anyone actually considering such a system in lieu of an A3000; there
                                                                          ^^^^^^
> is more to a computer purchase than system performance in the real world.
  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^  HMM ;-)
           =-=-=-=-=- lets quantify technical merit -=-=-=-=
> 
> >... but I can not understand your reticence to defend your creation with hard 
> >performance comparison results. Surely such results exist?
> 
> As I said, I don't sit around all day running benchmarks, and to my knowledge,
> none of the very large benchmark suites have been ported to AmigaOS.  The
> best Dhrystone 2.1 figure I've heard was something around 8000, other than
> that, I couldn't tell you.
> 
 And the above once more illustrates my point. You know very well
 that the Dhrystone benchmark is integer and character manipulation
 You have on occasion made this point yourself. Yet you post it here
 in response to an explicit request for floating point benchmark
 comparison results. ( the request is also somewhat more specific
 in that the comparisons are to be for *available* configurations,
 not ones in development )


-----------------------------------------------
  Reply-To:  Rostyslaw Jarema Lewyckyj
             urjlew@ecsvax.UUCP ,  urjlew@unc.bitnet
       or    urjlew@uncvm1.acs.unc.edu    (ARPA,SURA,NSF etc. internet)
       tel.  (919)-962-6501