top500 list (was: opteron VS Itanium 2)

Robert G. Brown rgb at phy.duke.edu
Mon Nov 17 12:05:35 EST 2003


On Mon, 17 Nov 2003, Dean Johnson wrote:

> Perhaps they should also have something like watts per gigaflop, or
> cubic feet occupied per gigaflop.

I actually think that it would be very interesting to plot watts/flop
over time, for integrated systems (not just "the CPU", but CPU and whole
box supporting).

My personal theory is that it is actually decreasing, on average,
because of interactions between Moore's Law scaling of CPU speed and CPU
power consumption and because of the cost of feeding the REST of the
system, which tends to hang nearly constant (effectively dividing it as
the relative power of the CPU is increased).  At any rate, as I run a
300 MHz Celeron side by side with a 2200 MHz Celeron at home, I don't
think the 2200 MHz Celeron eats 7+ times the power.  One day I'll
liberate my kill-a-watt and take it home and find out.

   rgb

Robert G. Brown	                       http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb at phy.duke.edu



_______________________________________________
Beowulf mailing list, Beowulf at beowulf.org
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf



More information about the Beowulf mailing list