On The Dangers Of Misapplying Moore's Law - Interesting Essay From Salon

[hatrack]

When will Moore's Law no longer hold? For 40 years, semiconductor manufacturers have successfully lived up to the challenge of doubling the number of transistors they can squeeze onto a chip of silicon every 18 to 24 months. This magic trick has bequeathed the world computing devices that are constantly pulling off the astonishing feat of delivering greater power at less cost. Occasionally, someone will predict that the party is about to end, but so far, no one has been right. As a consequence, culturally speaking, we expect that the iPod Nano or iPhone that hits the stores six months from now will have twice the memory and be cheaper than what's available today. This expectation, satisfied, has become practically an inalienable right.

And so we head relentlessly on to the next "process node." Right now, the bleeding edge of semiconductor technology is considered to be the 45 nm process node -- by which it is meant that the elements on a chip are about 45 nanometers apart, (although in reality, some are closer and some are farther.) Work is well under way figuring out how to achieve the next process node, 32 nm. But what happens after that? Electronic News offered a new twist on this age-old question last week, suggesting that while semiconductor manufacturers have a pretty good handle on how to achieve the next couple of generations, they are uncharacteristically at sea as to how to make the leap into even more nanoscopic territory.

EDIT

Metaphorically speaking, the end of Moore's Law would be quite a shock. Moore's Law inspires the kind of techno-utopianism that believes, almost as an act of faith, that humanity can innovate itself out of the messes it creates by sheer cleverness. Peak oil? Don't worry about it -- once Moore's Law starts working its magic on solar power, we'll have all the energy we need. World hunger in an era of drought and devastating climate change? No problem. Moore's Law applied to biotechnology tells us that we will keep redesigning plants to deliver ever greater yields under ever more drastic conditions.

There are some truths buried in there. Our fantastically exploding knowledge about the genetic structure of life would not be possible without cheap computing power. The application of semiconductor manufacturing economies of scale to solar power will bring down costs and make solar power a competitive source of electricity. But there's also a dangerous assumption built in; that we can keep this hustle going on forever. We're like those chip execs who are confident that they can to the next level, even if they have no idea what technologies they'll have to devise to get there. Oh, we'll think of something, we always have. But when you start measuring things in angstroms, isn't it possible that you're beginning to run out of room?

EDIT/END

http://www.salon.com/tech/htww/2007/07/27/moores_law/index.html

[phantom power]

1. Post-Soviet Russian computer scientists and numerical analysts were in high demand...

because the Soviet computer hardware was slower than American. They had compensated by becoming very very good at algorithms. We American computer scientists are a bit lazier, since we are used to riding Moore's Law.

So, the implication is that an end to Moore's Law would usher in a new era of people working harder at understanding algorithm optimization, parallelizing algorithms, etc.

Sometimes I dream of this new golden age. I always liked optimization, but it's sometimes hard to get people to pay for it in a climate dominated by Moore's Law.

[tinrobot]

2. When you can't go smaller, make more.

As the density of silicon nears it's limit, manufacturers will just start adding processors. We already have dual and quad core chips, and plans are in the works for even more cores.

[Dogmudgeon]

3. The next leap will have to be upward

We will have to start doing our heavy work in habitats in space. We will probably come to depend on them.

If we can not make that leap, there will be radical and painful retrenchment, and probably a dramatic reduction of population, too.

"Peak Oil" is the first of the resource-limitation crises we will face. Base-load primary energy will come next, and climate-driven agricultural collapse will probably be third. It's probable that these problems will all hit within a decade of each other. And even more, there are almost certainly going to be still other problems that aren't even on the radar.

We are not dealing with them very well.

The next generation of solar PV technology involves thin-film cadmium-tellurium semiconductors. Gallium and arsenic are central to LED manufacture. Mercury is used in quantity by many semiconductor processes. All of these substances are unbelievably toxic, and many of them cause horrifying sickness and death. Nobody seems to be concerned with toxic chemical control or process risk improvement -- some of the people I've spoken to are quite adamant that it poses little or no problem, even denying it. Yet they assume that industry will be able to increase semiconductor manufacture on the order of 1000 times, and do it quickly.

An obvious-but-ironic comparison can be made. I will leave it to your imagination to provide it.

Nanotechnology is also developing at a brisk pace. The big breakthrough in nanotech will be the replicating assembler. When Bill Joy wrote that the future does not need us, he was talking about just that. But by that time, it is possible that we will be on the way out.

--p!

[funflower]

4. It makes sense that chip capability growth potential is not infinite. eom