<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"
	xmlns:content="http://purl.org/rss/1.0/modules/content/"
	xmlns:dc="http://purl.org/dc/elements/1.1/"
	xmlns:atom="http://www.w3.org/2005/Atom"
	xmlns:sy="http://purl.org/rss/1.0/modules/syndication/"
		>
<channel>
	<title>Comments on: SiCortex&#8217;s low power, low latency super still on for a summer ship</title>
	<atom:link href="http://insidehpc.com/2007/06/25/sicortexs-low-power-low-latency-super-still-on-for-a-summer-ship/feed/" rel="self" type="application/rss+xml" />
	<link>http://insidehpc.com/2007/06/25/sicortexs-low-power-low-latency-super-still-on-for-a-summer-ship/</link>
	<description>HPC News Without the Noise for Supercomputing Professionals &#124; insideHPC</description>
	<lastBuildDate>Sun, 09 Jun 2013 01:54:13 +0000</lastBuildDate>
	<sy:updatePeriod>hourly</sy:updatePeriod>
	<sy:updateFrequency>1</sy:updateFrequency>
	<generator>http://wordpress.org/?v=3.1.1</generator>
	<item>
		<title>By: John Leidel</title>
		<link>http://insidehpc.com/2007/06/25/sicortexs-low-power-low-latency-super-still-on-for-a-summer-ship/#comment-42149</link>
		<dc:creator>John Leidel</dc:creator>
		<pubDate>Fri, 14 Mar 2008 13:52:25 +0000</pubDate>
		<guid isPermaLink="false">http://insidehpc.com/2007/06/25/sicortexs-low-power-low-latency-super-still-on-for-a-summer-ship/#comment-42149</guid>
		<description>Indeed... apologies for the erroneous numbers.</description>
		<content:encoded><![CDATA[<p>Indeed&#8230; apologies for the erroneous numbers.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: John Goodhue</title>
		<link>http://insidehpc.com/2007/06/25/sicortexs-low-power-low-latency-super-still-on-for-a-summer-ship/#comment-41989</link>
		<dc:creator>John Goodhue</dc:creator>
		<pubDate>Thu, 13 Mar 2008 18:55:25 +0000</pubDate>
		<guid isPermaLink="false">http://insidehpc.com/2007/06/25/sicortexs-low-power-low-latency-super-still-on-for-a-summer-ship/#comment-41989</guid>
		<description>John Leidel laid out the calculation better than I could, though I think he meant to say that the Intel configuration consumes 3 times the power of the equivalent SiCortex configuration.  (54.6kw/18kw = ~3).</description>
		<content:encoded><![CDATA[<p>John Leidel laid out the calculation better than I could, though I think he meant to say that the Intel configuration consumes 3 times the power of the equivalent SiCortex configuration.  (54.6kw/18kw = ~3).</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: John Leidel</title>
		<link>http://insidehpc.com/2007/06/25/sicortexs-low-power-low-latency-super-still-on-for-a-summer-ship/#comment-41869</link>
		<dc:creator>John Leidel</dc:creator>
		<pubDate>Thu, 13 Mar 2008 01:19:51 +0000</pubDate>
		<guid isPermaLink="false">http://insidehpc.com/2007/06/25/sicortexs-low-power-low-latency-super-still-on-for-a-summer-ship/#comment-41869</guid>
		<description>cstork... lets analyze things just abit.  Indeed, I agree with you that an current Intel [call it Clovertown] with the addition of its FBDIMMs will run at 400w.  However, when one adds all the board/card/disk peripherals and peg the cpu(s) on the machine, you can easily measure the power utilization at between 600 and 650w.  I&#039;ve done it... 

So, lets extrapolate this a bit.  
Given the following : Intel Clovertown, dual socket, quad core 2.0Ghz, 4 ops/clock.  
2.0Ghz * 4 * 8cores = 64Gflops peak.  Not bad.  Lets consider the peak power utilization to be 600w [machine is pegged].  
So, we have 64Gflops at 600w.  

In order to match the SiCortex machine, we would need exactly 91 nodes.  So, 91 nodes @ 600w = 54.6kw.  Ouch!
This is roughly 33% higher than the SiCortex.  

This general comparison is, of course, ignoring one&#039;s target workload, which is often the #1 driver in procurement activities.</description>
		<content:encoded><![CDATA[<p>cstork&#8230; lets analyze things just abit.  Indeed, I agree with you that an current Intel [call it Clovertown] with the addition of its FBDIMMs will run at 400w.  However, when one adds all the board/card/disk peripherals and peg the cpu(s) on the machine, you can easily measure the power utilization at between 600 and 650w.  I&#8217;ve done it&#8230; </p>
<p>So, lets extrapolate this a bit.<br />
Given the following : Intel Clovertown, dual socket, quad core 2.0Ghz, 4 ops/clock.<br />
2.0Ghz * 4 * 8cores = 64Gflops peak.  Not bad.  Lets consider the peak power utilization to be 600w [machine is pegged].<br />
So, we have 64Gflops at 600w.  </p>
<p>In order to match the SiCortex machine, we would need exactly 91 nodes.  So, 91 nodes @ 600w = 54.6kw.  Ouch!<br />
This is roughly 33% higher than the SiCortex.  </p>
<p>This general comparison is, of course, ignoring one&#8217;s target workload, which is often the #1 driver in procurement activities.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: cstork</title>
		<link>http://insidehpc.com/2007/06/25/sicortexs-low-power-low-latency-super-still-on-for-a-summer-ship/#comment-41249</link>
		<dc:creator>cstork</dc:creator>
		<pubDate>Mon, 10 Mar 2008 14:26:34 +0000</pubDate>
		<guid isPermaLink="false">http://insidehpc.com/2007/06/25/sicortexs-low-power-low-latency-super-still-on-for-a-summer-ship/#comment-41249</guid>
		<description>This doesn&#039;t seem so impressive.  5832 Gflops for 18 kilowatts?
 
Machines built on the latest Intel chips can do 100 Gflops for about 400 watts, 60 machines would be 24 kilowatts.</description>
		<content:encoded><![CDATA[<p>This doesn&#8217;t seem so impressive.  5832 Gflops for 18 kilowatts?</p>
<p>Machines built on the latest Intel chips can do 100 Gflops for about 400 watts, 60 machines would be 24 kilowatts.</p>
]]></content:encoded>
	</item>
</channel>
</rss>
