G E E K   P A G E    Issue 2.04 - April 1996

By Jef Raskin



Portable computers now have the speed, power and memory of the massive, multimillion-dollar machines that were state of the art a few decades ago. We all know that this is due to the microminiaturisation of silicon chips and other hardware. But another key factor has got smaller as well - and will have to get smaller still, especially if we ever want to make brain-sized, brainpower computers. That factor is one of the simplest properties of electricity: voltage.

Where vacuum-tube computers worked on dozens or hundreds of volts - enough to give you a good shock - most computers now run on a measly 5 volts, comparable to what you'd find in a four-cell flashlight. Newer microprocessors and associated chips run on 3.3 volts, and the voltage keeps dropping with each new generation. To see why and figure out how far, we can use our brains as a model.

I mentioned that higher voltages can be shocking; this is because your skin, basically an insulator, can conduct electricity if the electrons push hard enough. How hard electrons are trying to get through is what we mean by voltage. As lightning demonstrates, enough voltage will push electrons through the air, normally a good insulator. And through you, too, if you are in the way.

The higher the voltage, the greater the thickness of insulation electricity can punch through. That's why high-voltage power lines are so far apart and have such long insulators to hold them away from the metal "marching monster" towers. If the wires were closer together, the electricity might jump from one to another. Distance matters. For many common materials, we can list their insulating qualities in terms of how many volts it takes to force electrons through a certain thickness of the substance.

Typical values are on the order of a thousand volts per centimetre. So if an insulator is one-thousandth of a centimetre thick, then it might fail if it has as little as one volt across it. A typical flashlight battery puts out 1.5 volts, enough to fry so small a circuit. But the size of the electronic details on our chips is now dropping down into the submicron range. A micron is one-millionth of a metre, or a ten-thousandth of a centimetre, which means that if we place two conductors as close together as we possibly can with present commercial technology, we can't have much voltage between them.

There are two other factors that force us to use lower voltages on newer chips: speed and power. Information in a computer is transferred from one place to another along wires, and this information is usually coded in the form of rapidly changing voltages. One particular voltage level represents a binary one, a different level represents a zero. To change a voltage takes time - the smaller the change, the less the time. So a lower-voltage chip can run faster.

By power I do not mean computational power - of which the more the better - but electrical power, measured in watts. For a given chunk of material, the power of the electricity that passes through it - that is, the amount of heat generated in the material - goes up as the square of the voltage. Ouch. Computer designers already have problems with their machines running too hot. Lowering the voltage eases this problem - and helps extend battery life for portable computers.

Why not work, then, with millionths and billionths of volts? Because such tiny signal voltages can get lost in the electrical "noise" that abounds in our universe. A digital voltmetre can register a few millivolts just by touching your fingers. How low can we safely go? Let's look at a currently available and spectacularly superminiaturised computer - the human brain.

To pack one brainpower into something the size of our head, our biological cells have to be very small; we need trillions of them. (Compare this with the mere hundreds of millions of active units in today's computers.) The small size of brain cells, however, means that their voltages have to be very low. The problem is exacerbated by the fact that our cells are wet and are not good insulators. They have much lower breakdown voltages than air. All of this makes it clear why our nerves run in the millivolt range. This is just about the minimum that computer circuits can handle without being ruined by noise - at least with current technology.

In spite of the great improvement wrought by going to lower-voltage chips, designers have still not fully solved the problem of getting rid of the heat generated in the wires that bring power to the chip's various internal parts. The brain avoids this source of heat by bringing in its electrical power as juice - as a liquid. ("Juice" today is an informal synonym for electricity. Some day...)

The brain is bathed in fluids, and each cell uses chemical energy from those fluids to generate electricity on the spot. The fluids used to bring energy into the brain also act as coolants. In this regard, nature's method is a big win over today's human-designed circuits.

Given the efficiency of nature's slow design-and-test processes, we can confidently predict that when computers begin to rival the human brain in size and ability, they too will have signals measured in thousandths of a volt. Meanwhile, we can only stand back and narcissistically admire the wet computers in our heads, which allow us to invent devices that can project images and sounds over millions of miles, write and appreciate beautiful poetry and music and think of abhorrent concepts such as "ethnic cleansing." All it takes is a few millivolts in the right - or wrong - place.

Jef Raskin (jefraskin@aol.com) is best known for creating the Macintosh project at Apple. He now writes and consults from Pacifica, California.