Ternary (or even higher n-ary) storage is already widely used: multi-level cell (MLC) SSDs store more than one bit per cell by using multiple voltages (most commonly 4, i.e. 2 bits per cell). I believe Ethernet also uses 3 voltage levels.

Communication systems in general routinely have up to 1024 modulation symbols (32 voltage levels for both the sin and cos carrier).

http://en.wikipedia.org/wiki/Quadrature_amplitude_modulation...

Think hard drives - bit density = capacity

Let me elaborate on my post, since it has been misinterpreted. I was responding to the ideas in the article.

I of course understand the advantages of non-binary encodings in the many applications people mentioned, e.g. hard disks, flash memory, etc. Another common use is digital cable, which generally uses 256-QAM.

But let's look at the article itself, which was discussing ternary for general purpose CPU design. Here's a snippet:

   the first working ternary computer was ...
   designed by Nikolai P. Brusentsov and his
   colleagues at Moscow State University ...
   Some 50 machines were built between 1958
   and 1965. [It] operated on numbers composed
   of 18 ternary digits, or trits ...

I stand by what I first said. The above computers were probably wonderful academic toys, but otherwise were "a cure for no known disease".

Neither Intel nor anyone else will build a CPU with registers that store ternary numbers. Nor will they build cache memory that stores ternary numbers. Nor will there be any new ARM architecture forthcoming that switches from binary to ternary. Etc.