Probably it will be similar to Blue gene/P <br><a href="http://en.wikipedia.org/wiki/Blue_Gene#Blue_Gene.2FP" target="_blank">http://en.wikipedia.org/wiki/Blue_Gene#Blue_Gene.2FP</a><br><br>Each compute card could have a octa core processor and some memory chips.<br>
Like Blue Gene/P, each node card could have 16 compute cards and each card will be 2U high (but they will put two sideways like older Blue Genes).<br>So they can pack 100 cores in a 1U space without magic! :) <br><br>But cool this thing will now be easy.<br>
Probably they will use something like refrigerated doors (or walls).<br><br>
<br><div class="gmail_quote">2009/2/4 Robert G. Brown <span dir="ltr"><<a href="mailto:rgb@phy.duke.edu" target="_blank">rgb@phy.duke.edu</a>></span><br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<div>On Wed, 4 Feb 2009, Kilian CAVALOTTI wrote:<br>
<br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
Hi John,<br>
<br>
On Tuesday 03 February 2009 18:25:04 John Hearns wrote:<br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<a href="http://www.theregister.co.uk/2009/02/03/llnl_buys_ibm_supers/" target="_blank">http://www.theregister.co.uk/2009/02/03/llnl_buys_ibm_supers/</a><br>
<br>
I make this 400 cores per 1U rack unit. How is the counting being done<br>
here?<br>
</blockquote>
<br>
Even for the intermediate "development" system, Dawn, the article quotes<br>
150,000 cores in 36 racks. Which makes about 100 cores in a 1U space. That's a<br>
hell of a density...<br>
</blockquote>
<br></div>
Wow in spades. If the cores drew only 2W each, that would be a "normal"<br>
amount of heat for 1U and adds up to being ballpark 10KW per rack. If<br>
they draw 5W each, that's between 550W and 750W (the latter if there is<br>
a power supply in there with them, although I'm guessing that they'll<br>
provide power through rails and not via onboard power supplies at that<br>
density). 750W makes each rack burn something like 30 KW. And 5W is<br>
not a lot to drive a high-clock core and memory.<br>
<br>
One wonders how they are acheiving/planning to achieve it. Clock<br>
throttling? six to eight 16 core CPUs? An array of 800 MHz PDA CPUs?<br>
Some combination of the above? Something else entirely? I haven't<br>
looked into the power requirements of cores beyond dual and quad, but<br>
just running the bus and attached memory seems like it would make such a<br>
system hot, hot, hot. Or use slow, slow, slow processors (but a lot of<br>
them).<br>
<br>
The latter isn't a crazy idea, depending on the kind of task this<br>
faster-that-fastest system is supposed to be faster on. Some sort of<br>
massively SIMD decomposable problem with minimal nonlocal IPCs where the<br>
per-processor tasks are modest and nearly independent would get the<br>
near-linear scaling required to use up 1.6 million cores, and it would<br>
explain the 1 MB of memory per core. Consider each node as representing<br>
(say) 10,000 neurons and you've got a 16 billion neuron neural net with<br>
some sort of semilocal topology. Not bad, actually.<br>
<br>
rgb<div><br>
<br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
This article says 4096 cores per rack:<br>
<a href="http://www.engadget.com/2009/02/03/ibms-sequoia-20x-faster-than-the-worlds-" target="_blank">http://www.engadget.com/2009/02/03/ibms-sequoia-20x-faster-than-the-worlds-</a><br>
fastest-supercomputer/ That seems more rational - 93 racks adds up to 380928<br>
cores<br>
</blockquote>
<br>
An even weirdest number around this system is the amount of memory listed.<br>
1.6TB of main memory for 1.6 million cores, that's a mere 1 MB per core... I<br>
hope it's a P, not a T.<br>
<br>
Cheers,<br>
-- <br>
Kilian<br>
_______________________________________________<br>
Beowulf mailing list, <a href="mailto:Beowulf@beowulf.org" target="_blank">Beowulf@beowulf.org</a><br>
To change your subscription (digest mode or unsubscribe) visit <a href="http://www.beowulf.org/mailman/listinfo/beowulf" target="_blank">http://www.beowulf.org/mailman/listinfo/beowulf</a><br>
<br>
</blockquote>
<br></div><font color="#888888">
Robert G. Brown <a href="http://www.phy.duke.edu/%7Ergb/" target="_blank">http://www.phy.duke.edu/~rgb/</a><br>
Duke University Dept. of Physics, Box 90305<br>
Durham, N.C. 27708-0305<br>
Phone: 1-919-660-2567 Fax: 919-660-2525 <a href="mailto:email%3Argb@phy.duke.edu" target="_blank">email:rgb@phy.duke.edu</a></font><div><div></div><div><br>
<br>
<br>
_______________________________________________<br>
Beowulf mailing list, <a href="mailto:Beowulf@beowulf.org" target="_blank">Beowulf@beowulf.org</a><br>
To change your subscription (digest mode or unsubscribe) visit <a href="http://www.beowulf.org/mailman/listinfo/beowulf" target="_blank">http://www.beowulf.org/mailman/listinfo/beowulf</a><br>
</div></div></blockquote></div><br>