2010 September : Ron Riesenbach's Blog

To see a world in a grain of sand

Filed Under Tech

I had a moment of awe during these ‘days of awe’ … but not the religious kind.

When I was a grad student at the Technion in the mid 80s, I designed a custom image processor on a microchip whose transistor channels were about 5 micrometers wide ( 5 x 10 ^-6   = 5  millionths of a meter). With this technology, I could line up about 15 transistors end-to-end across the width of a human hair. That seems as impressive to me now as it did then.

Human Hair

A lock from a standard Homo Sapien

However a few days ago, I read a story about Intel’s recent announcement that it’s upcoming generation of processors will have transistors 32 nanometes wide, with a roadmap taking them to 22 and then 15 nanometers in just a couple of years. Fifteen nanometers is 15 billionths of a meter. That’s almost a thousand times smaller than the technology I was working with twenty years ago.

With Intel’s new fabrication technology, I could have lined up over 5,300 transistors end-to-end across the width of a human hair. In fact, fifteen nanometers is so small, that I would have had to take into consideration that I would only get about 60 atoms of silicon to work with in the transistor channel. Quantum effects start to make themselves felt at this scale.  I would have paid more attention in my quantum physics classes had I known that in only a few decades designers would bouncing electrons off of only a few dozen slippery atoms.

What gave me pause when I read the article was not just that the advanced fabrication technology I used for my Master’s degree had been eclipsed a thousand-fold. What I marveled at was the market pull for these ever-shrinking devices. To me, it is a complete surprise use that these ultra small, ultra cheep devices are now being used for. No one a few decades ago would dream that homes would have dozens of high-power processors in them and that middle-class teenagers would walk around with a half dozen powerful computers for their personal use (laptops, cell phones, music players, smart-chip credit cards, etc.). Today, my teen-aged kids would find it difficult to keep up their studies, their hobbies and their friendships without using personal processors and the computer networks that link them. In an incredibly short time, western economies and society have quickly adopted and become dependent on cheap CPUs, memory and networks.

In their book Connected , Nicholas Christakis and James Fowler coin an apt term for the species who’s behaviour is mediated by social connections — Homo dictyous — “network man” (from the Latin Homo for “human” and the Greek Dicty for “net”) . They use this idea in a slightly different context (as an alternative to economic-based analysis of human behaviour), but I think the term is apt in a more technological sense. We have become deeply dependent on our silicon devices in ways that we don’t even fully understand.

With Moore’s law , we can look forward to even smaller devices by the end of the decade. In fact, fifteen nanometers is just an order of magnitude larger than the size of a DNA molecule (about 2nm). So, it is not unreasonable to expect that transistors will be the size of DNA by 2020. They will also be extremely cheap, extremely low-power and extremely ubiquitous.

Imagine the possibilities.

Addendum: Things keep driving forward. In the 6-weeks since I wrote this blog. On Oct 29, 2010, Intel, Samsung, Toshiba announced a consortium aiming for 10nm chips by 2016.