Newsgroups: comp.sys.cbm
Path: utzoo!utgpu!jarvis.csri.toronto.edu!godzilla.eecg.toronto.edu!leblanc
From: leblanc@eecg.toronto.edu (Marcel LeBlanc)
Subject: Re: what is this chip(65C802)
Message-ID: <89Apr2.223029edt.2745@godzilla.eecg.toronto.edu>
Summary: More 808x bashing, compare machine cycle not clock speed
Keywords: 6502, 8088, DRAM
Organization: EECG, University of Toronto
References: <16510@cup.portal.com> <7734@killer.Dallas.TX.US> <2310@maccs.McMaster.CA>
Date: Sun, 2 Apr 89 22:30:16 EDT

In article <2310@maccs.McMaster.CA> cs3b3aj@maccs.UUCP (Stephen M. Dunn) writes:
>In article <7734@killer.Dallas.TX.US> elg@killer.Dallas.TX.US (Eric Green) writes:
>>in article <16510@cup.portal.com>, A-N-Onymouse@cup.portal.com (John - DeBert) says:
>>$ 8MHz is too fast for DRAM? Shhh! Don't tell my HP! It has DRAM and an 8MHz
>>$ clock. Plus a faster one for the alphagraphics display. :)
>>$  Seriously, one wonders how one may get DRAM to work with those so-called
>>$ turbo clones with clocks of at least 8MHz.
>>$  Also, I have seen 70ns DRAMS advertised from time to time. 
 ...
>>An 8080sux takes 4 clock cycles to read a byte from the bus. The 6502,
>>65c02, etc. take 1 clock cycle. Thus a 8 mhz 6502 would be the
>>equivalent of a 32mhz 8086-based machine. Considering that 80x6-based
>
>The problem with your argument is that you assume that everything on an
>8086 takes four times as long as on a 6502.  I'd like to see you get a
>6502 to do a 16-bit signed integer division in under 50 clock cycles

Since Eric was talking about memory access, there's nothing wrong with his
argument.  An 8088/86 takes 3 clock cycles per memory access, NO MATTER WHAT
INSTRUCTION is being executed!  The same is true of the Z80 and 8080.  Eric
stated 4 clocks/access because these processors need an extra cycle to
decode the instruction on a fetch.  The 6502 takes only 1 clock/access.

This is why Dave Haynie was saying that cheap DRAMs wouldn't be enough for
an 8MHz 6502.  An 8MHz XT on the other hand is much slower, so you can use
almost any speed DRAM.

>(which would be four times faster than an 8086), or do based indexed
>addressing on a pitiful CPU with two 8-bit index registers.  You're quite
>right about how dreadfully slow 8086 memory accesses are; however, make
>sure that you don't get everybody out there in NetLand thinking that they
>should sell their XT clones and buy a 64.

Standard clones (today) are faster than the original 4.77MHz that were
common when the C64 was younger.  When I bought my C128 (ages ago), I
tried running a simple benchmark on it, then compared with the results on a
clone that I had borrowed from work.  To make the test "fair", I tested
floating point addition, since neither machine had floating point hardware.
I used a simple "FOR" loop in BASIC (you would really have to compare the
two BASIC interpreters if you wanted to publish 'facts').  As expected, the
2MHz C128 was 40% FASTER than the 4.77MHz clone.

If you just compare clock speeds, then you might believe that a 33MHz 80386
is faster than a 10MHz R2000, 88000, or SPARC (?).  If I was just interested in
speed, I would use any of these at 10MHz before a 33MHz 80386.

Marcel A. LeBlanc	  | University of Toronto -- Toronto, Canada
leblanc@eecg.toronto.edu  | also: LMS Technologies Ltd, Fredericton, NB, Canada
-------------------------------------------------------------------------------
UUCP:	uunet!utai!eecg!leblanc    BITNET: leblanc@eecg.utoronto (may work)
ARPA:	leblanc%eecg.toronto.edu@relay.cs.net  CDNNET: <...>.toronto.cdn