top500 list (was: opteron VS Itanium 2)

Keith D. Underwood kdunder at sandia.gov
Mon Nov 17 13:04:45 EST 2003


> because rmax/rpeak as being a sort of "balance-like" measure.
> it's also scale-invariant, to the first order at least.
> 
> within the same category of hardware (say, desktop microprocessors and a 
> premium but off-the-shelf interconnect), rmax/rpeak is interesting, 
> since $/cpu are very roughly comparable.

But the rpeaks vary by a factor of 2 or more...

> > You can always put more
> > of them together with more money, right?
> 
> right, which is why I want to somehow regress scale out of the measure.

no - that was sarcasm.  At 10,000 processors it is hard enough to build
a box that will stay up long enough to do a useful amount of work with
apps that run across all of the nodes.  At 100,000 processors, today, it
is pretty close to impossible.  And that is IF you can get your app to
scale that well.  Big IF.  (yes, monte carlo simulations can probably
scale that high.  Yes, you could probably build fault tolerant monte
carlo simulations.  Yes, it would be nice to run something other than
monte carlo on the machine.)

						Keith




_______________________________________________
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