The Post-Silicon future

in hardware on 2014-07-12 21:05 (#3QF)

It's hard to understate the impact of the silicon chip and the advances wrought by Moore's law and steady research and development in miniaturizing transistors on silicon. But we're getting close to the limits, and getting beyond 9nm puts us at the limits of physics, it would seem. So what's next?

IBM is hoping it will be the first to find out. Not one to shy away from the big bets, IBM is putting $3B into researching the next step, including having a go at 7nm processes, new semiconducting materials like Gallium-Arsenide, and technologies like carbon nanotubes and graphene. At 7nm, the game changes significantly, and quantum physics begins to matter as much as traditional physics. Steve Torbak points out there's hope for technologies like racetrack memory and neuromorphic memory, too.

Or maybe, there's still room for improvement with what we've got. We're not done with Systems-on-a-Chip, after all, and DARPA has recently taken this approach to put an entire communications stack on a dime-sized chip.

[Ed. note: All I know is, to watch the next generation of silly cat videos, we're going to need a serious boost in hardware. /grin]

3 comments

Self-assembly polymers (Score: 2, Interesting)

by zafiro17@pipedot.org on 2014-07-12 21:13 (#2FN)

Interestingly, just spotted this rejoinder from the MIT Technology Review [bit of a paywall here although I read without paying]. I highly recommend reading this article though. Here's a clip.

http://www.technologyreview.com/news/528921/self-assembly-shows-promise-for-extending-moores-law/

A radical alternative to conventional lithography now looks increasingly viable. Known as directed self-assembly, it involves using solutions of compounds known as block copolymers that assemble themselves into regular structures. Block copolymers are made up of different units (the blocks) that prefer to be separate, like oil and water; left alone, these compounds typically produce swirling, fingerprint-like patterns. But if guided by a "pre-pattern" of chemical guides made with conventional lithography, the block copolymers produce lines and other regular patterns. Crucially, those final patterns can have much smaller details than those of the pre-pattern. A final pattern made in this way can then be used as a template for the chemical processes that etch features into a silicon wafer-the same process that is the end point of conventional lithography.

3D chips (Score: 2, Insightful)

by bryan@pipedot.org on 2014-07-12 22:54 (#2FV)

It's only a matter of time before needing to go in the 3rd dimension like they've already started doing with NAND flash. In fact, during the last few years, semiconductor companies have tested out new process nodes with NAND flash before atempting them with complex structures like CPU cores. With rows and rows of identical gates lined up in a perfect grid, this makes sense as it's probably a lot easier to spot defects and implement improvements.

So, if you want a preview of what CPU process nodes will be like in a few years, simply look at what NAND flash is doing today.

Gallium Arsinide isn't New (Score: 2, Informative)

by wildwombat@pipedot.org on 2014-07-14 04:35 (#2G8)

It hasn't been used in mainstream computers but it isn't new. It was used in the Cray 3 (even if only one was ever produced) and that was back in the late 80s and early 90s. They've actually got one of those chips at the Computer History Museum in Mountain View. It is also used in some satellite communication electronics.

Even if it gets continued development and turns out to be technically superior to Silicon on not sure how it will fare. If it is found to be environmentally harmful it could end up restricted under something like ROHS. Anybody know how toxic it is once its in its crystallized form?

Oh, and here is a cool video I came across when I googled for the Cray 3 link. Its Seymour Cray talking about Gallium Arsinide at the 1988 Supercomputing Conference.

Cheers,
--WW