Silicon Carbide Wafer

IBM has created transistors out of graphene that apparently can be clocked to frequencies well above 100 GHz.  If this is true, we might be seeing a new light at the end of the tunnel for “Moore’s Law”.  And believe me, it is quite dark after the next decade or so.  Intel had already previously announced a 22nm process, but with that we are getting within two orders of magnitude of the size of the Silicon atom (diameter of Si is .222 nm according to Wikipedia).

So, what does this mean for the fate of the transistor?

Simply put, as the transistor asks to get smaller and smaller (this has to be done to create smaller, more robust and power-efficient chips), we arrive at a theoretical limit at which the channel width of the transistor becomes so small that its crucial switch-like characteristics are no longer present.  Up until now, it was impossible to think of a single-atom transistor, but researchers at Helsinki University of Technology claim to have made a single-atom transistor, too.

The Single-Atom Transistor (Finland)

What do these two bits of good news mean?  Well, from the latter it appears that this is VERY GOOD news for Moore’s law, which predicts that the number of transistors on a chip will double every 18 months or so.  (It’s proven to be a very hard law to follow before Intel’s breakthroughs in the last few years.)

Yet without any further advancements in the fabrication process of  the transistor, we risk running into a dead-end soon.  And that would be VERY BAD for technological advancement.

So… what’s it going to be? Smaller transistors? or FASTER transistors?

Who knows.  For me, the smaller transistor seems to make more sense… it’s the way things have been going as of late.  The trend seems to be that the smaller the channel-width, the less power consumption there is.  On the other hand, single-atom transistors are the shorter dead-end road of the two, because a single atom is the limit (which we are very close to reaching now anyway). This is opposed to the theoretical limits of a clock-driven transistor, which are based around the speed of light.

We’ll just have to wait and see… but I’m only hoping that IBM’s breakthrough doesn’t revert us back to the single-CPU paradigm.  Getting all programmers to think in terms of parallel processing is difficult, but it will be necessary.  Many of us have learned the hard way, that 1 CPU can only be clocked so fast… adding an additional processing unit to a system is easy to do and gives two times the performance.  But increasing the clock speed by two times is way trickier, and doesn’t give the same results.