When it comes to new processor announcements, their manufacturers are always keen to point out that they are made on a smaller "nanometer process" or "process size" than a year ago, hence they are n-times more powerful and energy-efficient. All is well, but if you are the curious type, you might be wondering what a process size is, does it eat carrots, and how does it relate to the speed with which your smartphone runs games or drains its battery.
Besides, isn't it somewhat counter-intuitive that processors actually become smaller, yet they are more powerful, but less power hungry? After all, we're conditioned to believe that bigger also means stronger, and stronger means more power is required.
Don't worry, it's not that confusing in practice. We'll try explaining the idea of process size right now.
What's a nanometer?
In essence, a microprocessor is not much more than a bunch of layers made of different materials. Stacking them in particular ways yields us the tiny electronic components used by the processor to crunch numbers – such as transistors, resistors, and capacitors. These don't look like the ones in that broken tube TV uncle Joe threw in the bin eons ago. They are microscopic, and laid out on a grid of squares that act as On and Off switches. The distance between those processor components is measured in nanometers, which represents one billionth of a meter. The less that distance is, the more stuff you can fit on the chip.
Is that all there's to it?
No, because there are more ways in which decreasing the distance between components results in more efficient chips. Shrinking the microprocessor results in lower capacitance between the transistor terminals, which increases their switching frequency. And since the dynamic power a transistor consumes when switching electronic signals is directly proportional to capacitance, the transistors end up faster and less power-hungry!
Sweet, right? It gets better. Those smaller transistors need less voltage to turn on, so they are driven by lower voltages. And dynamic power loss is proportional to the square of the voltage. When you diminish the voltage needed to drive current through the transistors, you magically - no, mathematically - end up cutting power consumption. And the final factor that makes semiconductor makers push smaller process sizes is cost. The smaller a component is, the more of it you can fit in the wafers on which semiconductors are manufactured. Although smaller process sizes need more expensive equipment, the cost of investment is offset by the per-wafer cost.
But why does shrinking the process size take the industry two years on average?
Alas, nature has its ways to balance things in its favor, instead of ours. That's why those small, powerful, and efficient transistors from the paragraph above are more prone to leaking current. If you make semiconductors for a living, that's quite the wall to beat your head against. Voltage leaks happen in squares that are in an Off state, and result in the chip consuming power while it's doing nothing. In an ideal world, all those squares in the layer mask grid would be completely stable, but little electricity buggers like fluctuations, gradients, and diffusion get more troublesome as electronics become smaller.
So how small can these components ultimately get?
From the table below, you can see that the most popular and powerful mobile processors right now seem to be hanging at 20nm or 28nm. But the smallest process in commercial deployment is 14nm, established by Intel and used for its desktop and notebook CPUs. The company is targeting a 5nm process for chips in 2020, and in 2028, the industry expects to reach a 1nm process. This might end up as the limit of the current manufacturing technology we're using, and at that point, the industry will have to consider other developments and materials. Since we already had a transistor as tiny as a single atom in 2012, it's safe to say chip makers will find a way to continue bettering their craft.
|Processor||Process size (nm)|
|Qualcomm Snapdragon 810||20|
|Qualcomm Snapdragon 805||28|
|NVIDIA Tegra K1||28|
|Samsung Exynos 7 Octa||20|