5mn IC Fab the Limit?
The core technology underpinning modern Microprocessor design and fabrication has bearly changed since the first integrated Circuit was developed by Jack Kilby in 1958.
Today current state of the art IC fabrication is down to 5nm, with TSMC announcing a roadmap of 3nm by 2022. This technology is currently limited to simplier ICs such as RAM, current state of the art in more complex ICs are being led by AMD at 7nm. These fabrication sizes are impressive, however there is consensus that we are reaching the limits of photolithography. If you’re interested in understanding the techncial detail behind TSMC’s 5nm process then I recommend taking a look at this WikiChip article.
Although there’s excitement (may be hype?) around novel computing architectures such as Quantum, the reality is the vast majority of computing will still be on traditional processing architectures for some time to come. Therefore the challenge is how do we continue to deliver Moores Law?
Carbon Nanotube to the Rescue
This is where the technology of Carbon Nanotubes comes in. Carbon Nanotubes have very high levels of electrical conductivity over extremely small distances making them ideal as a building block for transistor circuits. However the design and fabrication processes are extremely challenging, in particular designing manufacturing processes that can be reliable and scaled up to enable the technology to be econmically viable for large volume IC production.
MIT made a breakthrough last year with a prototype CPU comprised of 14,000 Carbon Nanotube transistors (CNFETs). The key challenge they overcame was the instability in the fabrication process. They achieved this through developing circuit design software utilising machine learning to select the most likely robust transistator gate design.
One interesting aspects of MITs CPU, called RV16XNano, was the Instruction Set Architecture (ISA) of RISC-V. I’m planning an article on RISC-V in the near future.
The team at MIT are confident that we will be seeing commercial ICs built on Carbon Nanotube technology with the next five years. I, for one, believe this is the key transformation technology that will keep Moores Law true for the next 50 years.
In-Depth
If you’re interested in a deep dive on MITs work, a paper on the programme is published in Nature.