CCL: W:hardware for computational chemistry calculations

 Sent to CCL by: "Perry E. Metzger" [perry .. piermont.com]
 "Bill Ross ross ~ cgl.ucsf.edu" <owner-chemistry-#-ccl.net>
 writes:
 >> With time, for most problems,
 >> the needle points more and more often towards "get more RAM"
 than "get
 >> a faster disk". The obvious exception is stuff where getting
 things
 >> onto disk is mandatory (database servers), and where your working set
 >> is so large (right now past, say, 16G, soon much larger) that it is
 >> literally not going to be practical to get a machine with that much
 >> memory. Below that size, you owe it to yourself to do the
 calculation...
 >
 > I wonder if it's worth factoring in the power cost - don't know if
 > disk speed costs more power for the speedup than increased memory.
 You have reminded me of yet something else I neglected to
 mention. Clusters eat power, and once they've eaten the power, they
 turn it into heat, which eats more power and equipment to
 remove. You're very right to mention power considerations -- though
 even beyond the raw money involved in paying for the electricity, it
 costs money to modify a machine room to power big clusters, and it
 costs money to bring in sufficient air conditioning to remove the
 resultant heat, so the power budget becomes a serious consideration at
 times. One good thing is that if you overbuild once, machine rooms can
 often last a whole lot longer than the computers in them, but
 organizations planning for their first cluster often get seriously
 shocked by discoveries that play out like "what do you mean, there
 isn't enough power in the building!?"...
 Luckily, with time, computation per unit power tends to go down, but
 unluckily, we get greedier and greedier about the amount of
 computation we want to do. We also have this looming problem that in
 another 10 or 20 years, we're going to get very near the kT limit --
 for those that don't know, there is a minimum amount of power dictated
 by the laws of thermodynamics for an irreversible computation, and
 although we're still orders of magnitude above that limit, it is
 already clearly visible on the horizon...
 Perry