[Beowulf] Unaffordable Petascale ?

[Eugen Leitl]

On Mon, Apr 08, 2013 at 11:09:58PM +0000, Lux, Jim (337C) wrote:

> I don't see electricity costs rising by even an order of magnitude over the next few decades.  It's primarily driven by the energy cost to generate the electricity..

Unfortunately we're falling off the net energy cliff
http://www.resilience.org/stories/2008-09-14/net-energy-cliff
Oil peaked 2006 but the net energy decay hasn't started in
earnest before 2012.
 
> Crude oil is particularly volatile, and it's only gone up to $100/bbl vs about $20/bbl (inflation adjutsted) since 1945.  There have been some spikes (dec. 1979), but still , less than one order of magnitude.

History is not a good forecast of future in this point.
http://www.theoildrum.com/node/9249
 
> Coal prices are much less volatile. ($36/ton in 1949 to $32/ton in 2011, dipping as low as $19/ton in the early 2000s, and $24/ton in 1960s) (inflation adjusted).

Unfortunately, peak total fossil has been moved forward
to 2020 (peak coal was supposed to be 2030).
http://www.energywatchgroup.org/fileadmin/global/pdf/EWG-update2013_long_18_03_2013.pdf
 
> Recent advances in natural gas production have made it economic to produce an enormous amount of gas, which has always been cheap.  Oil is most rare: the organic matter has to have been at just the right temperature (about that of hot coffee) for long enough, and it has to be in a kind of rock that is porous enough to allow it to flow, and in a geology that allows it to be trapped (anticlines, faults, salt domes) for extraction.  Coal is quite common (too hot or cold and you still get coal, just of different grades), but still requires some heat/pressure.  Some coal is little more than dense dry peat (lignite, braunkohle)

Fracking is a flash in the pan, unfortunately. Read the numbers,
they're awful.
 
> Methane is everywhere. It's what pretty much all organic matter turns into if it doesn't turn into coal or oil.  Until the last 20 years or so, it was either an inconvenience (pressurizing a dry hole when looking for oil, making coal mines hazardous, causing department store basements to explode in Los Angeles), or a convenience (pressurizing the oil so you don't need to pump it) or it was in a formation that isn't porous enough to get decent flow rates.  These days, though, there's a lot of ways to get it out of the ground, there's more infrastructure to move it around, and it's a heck of a lot better fuel than just about anything for something on land (no problems with heavy metals, not too much problem with sulfur).  I was surprised recently to see that that natural gas is about to overtake coal as the leading source of electricity in the US (1.2 PWh for gas, 1.5 PWh for coal out of about 4PWh total)

Look forward a decade, the bonanza is not there. US is setting itself
up for a hard fall.
 
> Since people are running PFLOPS today.. and can afford it, I can't see an ExaFLOPS being out of the question.

We're many orders of magnitude away from physical limits (spintronics should
make a very dramatic impact already), and of course if you never have to
erase bits (reversible computing) you can even go quite a bit lower.