Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!csd4.milw.wisc.edu!lll-winken!uunet!dg!rec
From: rec@dg.dg.com (Robert Cousins)
Newsgroups: comp.arch
Subject: Re: Criteria for comparing RISC processors
Message-ID: <157@dg.dg.com>
Date: 3 May 89 14:35:42 GMT
References: <2368@ogccse.ogc.edu> <1464@cfa.cfa.harvard.EDU> <141@dg.dg.com> <18120@winchester.mips.COM> <144@dg.dg.com> <18316@winchester.mips.COM> <147@dg.dg.com> <18653@winchester.mips.COM> <102441@sun.Eng.Sun.COM>
Reply-To: rec@dg.UUCP (Robert Cousins)
Organization: Data General, Westboro, MA.
Lines: 64

In article <102441@sun.Eng.Sun.COM> jek3@sun.UUCP (Joseph Kowalski) writes:
>In article <18653@winchester.mips.COM> mash@mips.COM (John Mashey) writes:
>>In article <147@dg.dg.com> rec@dg.UUCP (Robert Cousins) writes:
>>.....
>>>                                       Note that compilers for the 88K
>>>produce code RIGHT NOW with performance (Mhz for Mhz) equal to the MIPS 
>>-----------------------------^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>>>compiler, which has been in development for years. 
>
>I fail to see the preoccupation with 'Mhz for Mhz'. I certainly would like
>to believe that one can design a simple state machine (like a D flip-flop)
>to run at a higher clockrate than a DSP or CPU chip (like 88K or R3000).
>If an architect chooses to implement an instruction set to aid compiler
>designers/performance at the expense of clockrate, that's his design
>decision. I certainly don't want to see 33 Mhz 88K compared against an
>R2000 at 8 Mhz (or other sadism), but I feel that the following is a much
>more meaningful comparison than 'Mhz for Mhz':
>
>	Compare performance on the highest performing currently available
>	implementation of an architecture in a given technology
>
>I realize that 'Mhz for Mhz' has more meaning in the RISC, ONE INSTRUCTION PER
>CLOCK world than in the CISC world, but even this is doubtful.
>
>Joe Kowalski (No fancy signature - My opinions are obviously my own - I tried,
>	      but I couldn't give them away!)

Your point is well taken.  By saying "Mhz for Mhz," I was attempting to make
sure that reasonable machines are being compared.  There is a penchant in
this world for people to compare 25 Mhz MIPS machines with 16 Mhz 88Ks.
Another way to put it is to compare machines dollar for dollar.  It is
not suprising that a $40K machine should compare well in performance to a
machine selling for under $10K.  We can talk $/Dhrystone MIP.  We can
talk $/Mhz or we can flat out match high end to high end.  However,
each of these has its disadvantages.  

Dollar for dollar comparisons don't apply when there is no overlap
	between the machines.  For example, comparing a Cray with
	a PC. :-)

High end to high end suffers from the same problem:  one person's high
	end is another person's low end.  This is especially true in
	the 88K world where a limited number of products are currently
	announced.  If you want to base this upon announments of planned
	products, then the world changes again.  It was stated at the
	AViiON announcement that there will be a 500 MIPS quad ECL 88K product. 
	Should this be brought into the current discussion?  Probably
	not.

Low end to low end is somewhat more fair.  In this space, the design
	goals are more similar and therefore it is often easier
	to make a direct comparison.  However, the Cray vs PC problem
	still applies.

$/MIP is the most fair since it more readily allows people to judge
	in an apples-for-apples equation.  When one looks at a machine
	it should be possible to measure value (at least for a 
	known group of problems) in relation to this.
 
Comments?

Robert Cousins

Speaking for myself alone.