<rdf:RDF
    xmlns:rdf='http://www.w3.org/1999/02/22-rdf-syntax-ns#'
    xmlns:s='http://snipsnap.org/rdf/snip-schema#'
    xml:base='http://www.peerbox.com:8668/rdf'>
    <s:Snip rdf:about='http://www.peerbox.com:8668/rdf#start/2007-01-05/1'
         s:name='start/2007-01-05/1'
         s:cUser='kgr'
         s:oUser='kgr'
         s:mUser='kgr'>
        <s:content>1 3500 Core CPU? {anchor:3500 Core CPU?}&#xD;&#xA;According to this {link:article|http://www.zdnetasia.com/news/hardware/0,39042972,61979080,00.htm}, IBM’s next generation Power6 processor will have 700 million transistors and Sun’s next generation Niagra2 processor will have 500 million transistors.  That’s a staggering number of transistors (at least I’m staggered).  By way of comparison, the Motorola 68000 which powered the original Macintoshes, Amigas, and Sun workstations, had, coincidentally enough, only 68k transistors. An ARM RISC processor can be had for as little as 30-40k transistors.  If you used half of the transistor count for cache then this would still leave you with room for 3,500 100k-transitor processors or maybe even 35,000 simple 10k-transistor stack-based processors.   If you could find a good way to program and co-ordinate all of the little processors then such a design would certainly give you much more processing power than a small number of monolithic cores using the same transistor count.   Massive core counts could also help to increase chip yields.  Even a chip with dozens of defects could still be viable if you just masked out the broken processors and memory regions in firmware.   A single core with ten defects is completely useless whereas a 3,490 core CPU is practically indistinguishable from a 3,500 core CPU.  Most of today’s software wouldn’t run very well on such an architecture but if you consider that they are predicting that the average desktop computer will have 100 cores within a decade, then this might change.  Once you have to rewrite your software to work with a hundred cores anyway, then it may not be that much extra effort to make it work with a thousand or more.</s:content>
        <s:mTime>2007-01-05 15:44:59.618</s:mTime>
        <s:cTime>2007-01-05 13:06:38.049</s:cTime>
        <s:comments
             rdf:type='http://www.w3.org/1999/02/22-rdf-syntax-ns#Bag'/>
        <s:snipLinks>
            <rdf:Bag>
                <rdf:li rdf:resource='http://www.peerbox.com:8668/rdf#2007'/>
                <rdf:li rdf:resource='#kgr'/>
                <rdf:li rdf:resource='http://www.peerbox.com:8668/rdf#Things to Learn'/>
            </rdf:Bag>
        </s:snipLinks>
        <s:attachments
             rdf:type='http://www.w3.org/1999/02/22-rdf-syntax-ns#Bag'/>
    </s:Snip>
</rdf:RDF>
