<br><br><div class="gmail_quote">On Thu, Jul 15, 2010 at 8:12 PM, JD <span dir="ltr"><<a href="mailto:jd1008@gmail.com">jd1008@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
<div class="im">.</div><div><div class="h5">
><br>
</div></div>"There he goes again"<br>
I did not choose any position. I merely stated a fact and you concocted<br>
from that, that I am taking a position on the issue.<br>
I am an old timer by the average age of this list's subscribers. I was<br>
around when the pdp 8 and pdp 11 were used to teach<br>
assembly language programming.<br></blockquote><div><br></div><div>Quoting you:</div><div><br></div><div>"<span class="Apple-style-span" style="font-family: arial, sans-serif; font-size: 12.8601px; border-collapse: collapse; ">This is the kind of technology that is needed to solve huge mathematical</span></div>
<div><span class="Apple-style-span" style="font-family: arial, sans-serif; font-size: 12.8601px; border-collapse: collapse; ">problems."</span></div><div><br></div><div><span class="Apple-style-span" style="font-family: arial, sans-serif; font-size: 12.8601px; border-collapse: collapse; "></span>Highly questionable. Mostly wrong. Ok?</div>
<div><br></div><div>Bandwidth. Bandwidth. Bandwidth.</div><div><br></div><div>And it's a big problem for GPU's which are really, really good at pounding the living daylights out of small datasets, but of marginal usefulness for "huge mathematical problems," except for a small subset of such problems, like linpack. I've argued this over and over again publicly. People who actually control budgets already know what I have to say. The most interesting problems are bandwidth-bound, not compute-bound.</div>
<div><br></div><div>If I could have a Fermi, would I? You bet, because I want to play with one. Do I expect it to open up new ground that in any way reflects naive measures of its computational power? No, I don't.</div>
<div><br></div><div>Robert.</div><div><br></div></div>