Over half a century ago Gordon Moore, co-founder of Intel, postulated that the number of transistors in a circuit would double every 18 months. In the computer world, this assertion has become known as Moore’s Law and has, for the most part, held true. This happens because we’ve become exceedingly good at figuring out ways to shrink the size of transistors.
The CPU in your phone, computer, tablet, TV, car, whatever, contains billions of transistors that work in concert to make the CPU do what the CPU does. It’s the brains of the device, without our technology wouldn’t work. And it’s the single best component to improve if you want to increase the power or efficiency of your device.
Taiwan Semi-Conductor is one of the world's largest chip fabricators. The firm doesn't design chips, it creates the process of making the chips and other companies that design their own chips pay TSMC to make them. TSMC recently moved to a 5 nanometer (nm) process, which Apple is utilizing for its new chips. Apple recently announced a new chip that will be going into some of its computers. This is a departure from Intel, who struggled to shrink processors below 14nm (Intel also makes chips instead of paying another company to make them).
I should point out that these numbers don’t directly translate to the size of transistors and that each company has its own ‘definition’ of what those number means. For instance, Intel’s 14nm process is closer to 10nm at some of the competitors. But for the sake of brevity, we’ll use these numbers.
To give you an idea of how small this is. The width of a human hair is about 50 micrometers. 50 micrometers is 50,000 nanometers. So take a human hair and slice it, widthwise, into 10,000 equal parts and that’s roughly the size of a transistor in these new chips. Another way to look at it is the diameter of DNA is about 2nm.
These sizes are tiny. It’s an incredibly difficult process and one can see why it is taking Intel time to shrink its process. There are several reasons that shrinking transistor size creates more power and efficiency, but they’re fairly technical and gist of it is that the smaller the transistor the less power it requires. It’s not much difference per transistor, but when you multiply that tiny difference by the billions of transistors in the CPU it adds up to a huge gain.
And that leads us to another reason that shrinking transistors leads to power gains in the CPU.
When TSMC shrank its process from 7nm to 5nm, designers reduced the size by about 30%. That means in the same area you can now fit about 30% more transistors. More transistors equals more processing power. It also gives you flexibility. Now you can decide to keep the processing power roughly the same while decreasing the overall size of the CPU and reducing the power consumption greatly, or you can decide to keep the CPU the same size while increasing the processing power and reducing power consumption slightly, or some combination of the two.
The plans to shrink the size down to 3nm or 2nm are well underway. There is a physical limit to how small we can make these transistors, at that point we’ll have to figure something else out. But if history is anything to go by, I’m sure we will.