In 1965, Intel cofounder Gordon Moore first observed that the number of transistors that could fit onto a dense integrated circuit (or chip) doubled each year. In 1975, Moore adjusted this "Law" to approximately every two years. Since then, computing power has expanded at an astounding rate.
Is Moore's Law Dead?
For instance, the Apple 1, which premiered in 1976, featured a 1MHz processor and 4 kilobytes of RAM. Today, processor speeds are measured into the multiple-gigahertz range with multiple cores, multiple gigabytes of RAM, and accessories to process data at a speed that would have been mind-boggling not long ago.
These advances also allow the miniaturization of modern smartphones and any number of other devices. For the last decade—perhaps even longer—however, we've been hearing that Moore's Law is dead, dying, or on life support.
Carbon Nanotube Processor: The 5nm Transistor
If not the letter of the law, it's at least in the spirit of computing power advancing exponentially for years to come. Here's why:
- Regardless of how many more transistors we can fit onto a chip, engineers continue to find novel ways to optimize and take advantage of what's available. Consider, for instance, the rise of cloud computing for adding remote computing resources on-demand, and the implementation of devices like the Google Coral Edge TPU Accelerator for specific tasks.
- We are still finding ways to fit more transistors onto a chip. Per a 2014 article from MIT Technology Review, to stay true to Moore's Law, chips would have to have features in the 5-nanometer range to keep up. That's a staggering demand, considering silicon atoms themselves are only around .2nm. However, here we are in early 2020, and limited production of 5nm trials has been completed, with mass production slated to begin this year.
- As predicted in the MIT article, chips made using carbon nanotube transistors should be available in 2020. While it doesn't look like you'll be able to buy a "carbon computer" this year, researchers at MIT made excellent progress on this technology in 2019, creating a 16-bit microprocessor consisting of more than 14,000 carbon nanotube field-effect transistors (CNTFETs).
While this 14,000 CNTFET, RISC-based processor isn't ready to compete with modern silicon chips, it's still an amazing accomplishment. The CNTFET transistor itself wasn't demonstrated until 1998, and in 2013, it was a significant achievement when researchers were able to make a nanotube processor using just 178 of these transistors.
CNTFET Processor: Is Moore's Law Still True?
If we're looking at transistors on a chip doubling every two years per Moore's (current) Law, the six years between the 2013 and 2019 chip would predict only a 712 CNTFET processor. Doubling this rate every year would result in 11,392 transistors on a chip. Therefore, CNTFET advancement has exceeded even Moore's early prediction, an exciting development if this trend continues.
One aspect to overcome is that these chips typically need to be 99.999999% pure to function properly. This is impossible under today's processes; however, researchers were able to design around this obstacle for the 2019 RISC CNTFET processor, allowing them to make do with only 99.99% purity. Processes have also been developed to mitigate a large number of these defects directly. We can expect further advances as these techniques evolve, hopefully clearing the way for mainstream carbon-based computing.
Carbon Nanotubes vs. Silicon: More Efficient
More computing power generally equates to more power usage and challenges for heat dissipation. As it turns out, research indicates that CNTFET technology may be an order of magnitude more energy-efficient than its silicon counterparts. Carbon nanotubes are excellent at dissipating the heat they produce, a problem that plagues silicon-based computers as processors are clocked at higher and higher speeds to increase performance.
Moore's Law is Still Valid
So, it appears that Moore's Law isn't actually dead and that 5nm silicon will be a bridge to better computing until CNTFET chips arrive en masse. Perhaps we—or the next generation—will still be talking about the impending doom of computer advancement 30 years from now as we await even more advanced technology, such as quantum computing, to satiate our ever-present need for better and smaller devices.
And if that's not the case, it isn't as if Moore's Law hasn't been modified before!