IBM Reports Carbon Nanotube Chip Breakthrough 73
First time accepted submitter yawaramin writes "IBM has apparently made a breakthrough in arranging carbon nanotubes into the logic gates necessary to make a chip. This should help miniaturize and speed up processors beyond what today's silicon-based technologies are capable of. The article notes though that perfecting the carbon nanotube technology could take up the rest of this decade."
Re:citation? (Score:3, Interesting)
premature optimization postponed (Score:5, Interesting)
Actually, no. Micro-architecture could continue to evolve without die shrinks (likely toward a proliferation of specialized units) and software could also evolve. Probably both for a decade or so, before the shrink stall becomes a fed stall. A feature of Moore's Law rarely expressed is that software lags architecture, and architecture lags die size.
I realized a long time ago that if I could gain a 50% speed increase by rewriting a critical application loop in assembly language, it generally wasn't worth the bother. The next processor architecture would mess up you clever clock-count calculations. The effort was almost always better invested in satisfying feature demand as PCs became more capable. Not only does the architectures improve, but so does the cleverness of your compiler (not including your hand-polished asm). If the software people actually knew that die shrinks were a thing of the past, it would make sense to be more aggressive in the choice of algorithms and execution regimes. They might even be well paid to indulge in premature optimizations postponed, since this would become the main avenue to sustaining performance gains.
There might be more pressure to bet on the right horse, which could thin the herd. Competence gradients tend to have this effect.