Moore's Law Slow Death
Moore's Law has proved constant in the semiconductor industry, providing hope for more powerful computers. The law states that the number of transistors on a silicon chip doubles every 18 months, hence producing devices twice as powerful (typically). If a computer is slow today, never fear because just beyond the horizon is a new processor that could speed up computing significantly.
Well, Moore’s Law may finally come to an end in 2020 which means beyond the horizon may be a processor that is only slightly better or cheaper. So how will modern computing keep up with power demand? What alternatives exist for such a computational problem?
Some solutions seem rather simple while others are more exotic in nature. For example, to squeeze the most out of silicon may require a blast from the past with programmers aiming for optimization above all else. Object orientated programming is great for program creation but uses large amounts of resources—so programmers may shift to using assembly for routines that require speed. More efficient algorithms could also help with CPU efficiency as well as freeing up CPU resources for other operations.
Hardware designers, meanwhile, may look at implementing specialized integrated circuits to free up common tasks found on CPUs such as floating point operations and encryption. Other approaches could be the increased use of multiple CPUs on motherboards as most computers on the market still only have one CPU socket.
Transistors can only get so small
An exotic approach could include the implementation of quantum computers to help with algorithms, including encryption and analytical tasks. But such solutions are still being developed and require high-tech scientific equipment, making them largely impossible in the commercial sense.
Another solution may be a more dramatic take on computational problems by developing computers like brains or using brains, themselves, to perform computational tasks. No matter how you look at it, something needs to be done soon to improve the performance of computers and remove our dependency on more powerful silicon-based devices.
IBM’s Nanoscale Solution
Carbon nanotubes are not exactly the stuff of science fiction but rather science fact. These tiny tubes are a big contender in the “material for the future” contest because they are super strong, very conductive, and—above all else—very tiny. These structures, which are only one atom thick, consist of carbon atoms covalently bonded in a tube-like structure which have seen practical use already in applications including atomic force microscope probes, composite materials, and even bicycles.
However, their electrical characteristics are yet to be exploited and this is what IBM has in mind for chips of the future.
A visualization of carbon nanotube structures. Image courtesy of Michael Ströck [CC-BY-SA]
The problem with carbon nanotubes is that, while individual tubes are fantastic, they are not the easiest material to manipulate and turn into useful structures. When creating the structures, the carbon tube mixture consists of different lengths of tubes with different conductive properties (some metallic and some semiconducting). This is not ideal for the creation of a semiconductor device. So researchers were forced to create the required structures manually.
IBM later discovered that, if the right voltage was applied to the mixture, the metallic carbon nanotubes incinerated, leaving behind semiconducting tubes. While semiconducting tubes can now be separated, there is still the problem of structural arrangement. Attaching nanotubes together was thought to be nearly impossible but researchers at IBM again proved this wrong by attaching molybdenum to the ends of the tubes which can be used to “glue” tubes together. The inclusion of the molybdenum does not impact on the electrical characteristic of the nanotubes which is vital for functional electronic devices.
Nanostructures made by IBM. Image courtesy of IBM Research
The latest hurdle that has been solved by IBM is the structural arrangement of the tubes. Silicon devices are easily constructed (comparatively) because they are made from a slab of silicon which is treated with chemicals, exposed to light, layered with materials, and sliced into pieces.
Carbon nanotubes, however, cannot be manipulated in the same way using current manufacturing techniques. Scientists from IBM describe the process of carbon nanotube manipulation as more like growing crystals in which chemicals are used to alter the final structure.
Nanotubes are strong, light, and exhibit excellent structural properties. This is important for any semiconductor material but one of the most important characteristics is their size.
Currently, transistors with feature sizes as small as 10nm are being used in commercial designs but it is expected that commercial designs using sizes smaller than 7nm will be improbable. While transistor feature sizes as small as 1nm have been successfully made, the use of such transistors in a scaled-up commercial design may be too expensive to produce (for the computational power they offer).
This is where carbon nanotubes come in. The width of a carbon nanotube can be as small as 1nm across which already makes then a 10th of the size of current transistor features. On top of that, carbon nanotube structures are reliable and strong, unlike silicon transistors which are a hit and miss production operation.
It is these properties that make carbon nanotubes the future chip making material that could see faster processors that use less energy and dissipate less heat. Such processors would not just see use in domestic environments for their processing capability but also in large computational environments including data centers which rely on devices generating as little heat as possible.
IBM’s nanotubes might still be in their infancy, with a real device yet to be constructed, but there is no doubt that carbon nanotubes could potentially replace or be used in parallel with silicon devices. Regardless, Moore’s Law is likely coming to an end so we as engineers need to band together to find a solution that can continue the trend of more powerful computing.