Comments on: Highly Parallel HPC: Ants vs Horses http://www.linux-mag.com/id/7756/ Open Source, Open Standards Sat, 05 Oct 2013 13:48:18 +0000 hourly 1 http://wordpress.org/?v=3.1 By: dragansah http://www.linux-mag.com/id/7756/#comment-8225 dragansah Wed, 30 Nov -0001 00:00:00 +0000 http://www.linux-mag.com/id/7756/#comment-8225 <p>I think it should be 1/RT.</p> <p>Nice post </p> I think it should be 1/RT.

Nice post

]]> By: vassago http://www.linux-mag.com/id/7756/#comment-8226 vassago Wed, 30 Nov -0001 00:00:00 +0000 http://www.linux-mag.com/id/7756/#comment-8226 <p>Any chance we can get a working link on the \"Performance Considerations for I/O-Dominant Applications on Parallel Computers\"?, all excited and 404 :). Thanks. </p> Any chance we can get a working link on the \”Performance Considerations for I/O-Dominant Applications on Parallel Computers\”?, all excited and 404 :). Thanks.

]]> By: bugmenot3 http://www.linux-mag.com/id/7756/#comment-8227 bugmenot3 Wed, 30 Nov -0001 00:00:00 +0000 http://www.linux-mag.com/id/7756/#comment-8227 <p>vassago, that article can also be found here: http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.45.4406</p> <p>CiteSeer keeps articles cached in PDF and Postscript format :) </p> vassago, that article can also be found here: http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.45.4406

CiteSeer keeps articles cached in PDF and Postscript format :)

]]> By: vassago http://www.linux-mag.com/id/7756/#comment-8228 vassago Wed, 30 Nov -0001 00:00:00 +0000 http://www.linux-mag.com/id/7756/#comment-8228 <p>I totally just had a jfgi moment huh? Wow, too many things at once :). Thanks! </p> I totally just had a jfgi moment huh? Wow, too many things at once :). Thanks!

]]> By: maddognh http://www.linux-mag.com/id/7756/#comment-8229 maddognh Wed, 30 Nov -0001 00:00:00 +0000 http://www.linux-mag.com/id/7756/#comment-8229 <p>Interesting article, but it begs the question:</p> <p>Is it more energy efficient to use two Cray-1s (slower processors) than four faster Intel Atom processors? What about Cell processors?</p> <p>I agree that there are typically diminishing returns on parallelization for a single problem, but I think there are a lot of considerations left out of the equation. </p> Interesting article, but it begs the question:

Is it more energy efficient to use two Cray-1s (slower processors) than four faster Intel Atom processors? What about Cell processors?

I agree that there are typically diminishing returns on parallelization for a single problem, but I think there are a lot of considerations left out of the equation.

]]> By: deadline http://www.linux-mag.com/id/7756/#comment-8230 deadline Wed, 30 Nov -0001 00:00:00 +0000 http://www.linux-mag.com/id/7756/#comment-8230 <p>Fixed the link and the parallel resistance formula. </p> <p>Mad Dog, </p> <p>I think it is best considered in terms of, \"with a given CPU family\" it is more power efficient to run multiple processors at a slower clock than one processor at a higher clock. At least that is how I interpret the point Pfister made. I could have been more clear on that point. Of course there is much more to the idea because the argument assumes perfect parallelism (no overhead), but the point is still interesting. I think SiCortex was based on this idea. </p> <p>Thanks for the feedback! </p> Fixed the link and the parallel resistance formula.

Mad Dog,

I think it is best considered in terms of, \”with a given CPU family\” it is more power efficient to run multiple processors at a slower clock than one processor at a higher clock. At least that is how I interpret the point Pfister made. I could have been more clear on that point. Of course there is much more to the idea because the argument assumes perfect parallelism (no overhead), but the point is still interesting. I think SiCortex was based on this idea.

Thanks for the feedback!

]]>