While supercomputers may not be more efficient than the human brain, we still have managed to create machines that are more powerful. The Sunway TaihuLight is currently the fastest supercomputer in existence, measuring at 125 petaFLOPS. Basically, a “FLOPS” is a way to measure processing speed, one FLOPS is one floating-point arithmetic calculation per second. A petaFLOPS is 1015 FLOPS...that means this supercomputer can perform over 125,000,000,000,000,000 complex arithmetic calculations per second. Goddamn, meanwhile it took me about 15 seconds today to figure out that 62 + 3 is the same thing as 6 * 6 + 3.
What makes the Sunway TaihuLight’s impressive processing power even more unbelievable is the fact that supercomputers cracked the petaFLOPS barrier less than a decade ago, in 2008. Meaning that the processing power of the most powerful supercomputer went up by over 12,500% in just nine years. But, there has to be a limit...right? Of course there has to be a limit regarding how quickly a computer can process information.
The first major thing holding modern computers back is the transistor. You see, the transistor is a physical component that allows the computer to process information. So the more you can fit within a circuit, the more processes you can accomplish. For the past few decades, the size of the transistor has shrunk immensely. The very first usable transistor created in 1947 could fit in someone’s hand (about 9 cm) and now the smallest is about 1 nm.
Many scientists thought creating a transistor smaller than 5 nm was going to be impossible due to quantum tunneling and now they are stuck again with trying to shrink the transistor even more.
But the size of the transistor is not the only thing that determines the speed limit of the computer. In fact, we may not even have to worry about shrinking the transistor anymore with quantum computers becoming the norm.
Instead, Seth Lloyd from the MIT Department of Mechanical Engineering states, “In particular, the speed with which a physical device can process information is limited by its energy and the amount of information that it can process is limited by the number of degrees of freedom it possesses”. What does that all mean? It basically means that there is a universal limit to how many degrees of freedom a computer may possess, thus there is a finite amount of energy that can be transferred per second. Electrical energy must be used in order to process data, which is why computers actually have to be plugged in to work.
Seth Lloyd and his team go over all of the specific equations here, but the gist of the math is that by using E = mc2, where m is the mass of the computer and c is the speed of light, the average energy (E) that a computer uses can be determined. This is assuming ideal conditions. Then, the number of operations a computer could process by second is determined by taking the average energy (E) doubled, and dividing it by π and h, where h is Planck’s reduced constant. So, the equation to find the number of operations per second is # = 2E/πh. Basically, the larger a computer is, the more energy it can use per second, meaning a larger amount of operations can take place per second.
Using the above equations, you can find that a one kilogram computer, under circumstances where processing speeds are equal to the speed of light, can process up to 1.36 x 1050 operations per second. So the world’s fastest supercomputer, the Sunway TaihuLight, is about 1.36 x 1035 times slower than the theoretical limit of a computer that is one kilogram. By the way, the Sunway TaihuLight is much more massive than one kilogram, so we still have quite a ways to go.
But let’s get crazy. Say we could build a computer that is the mass of the entire observable universe (1053 kg). We would get a computer that has a limit of 5.426 x 10103 operations per second. That is more operations per second than the number of atoms in the universe.
So yeah, we have just touched the tip of the iceberg so far. Modern day supercomputers are fast, but on the grand scale of things, it is possible to go much, much faster. Realistically however, we will never hit the computational speed limit. It should be impossible for computers to ever process the things at the speed of light and unless we plan on building a computer the mass of the entire Earth any time soon, it looks like we will have to stick to using computers that only process 125,000,000,000,000,000 per second. What a bummer.