An article published in the journal “Nature” describes the creation of a new architecture for next-generation computers. A Stanford University and MIT team led by Max Shulaker, assistant professor at MIT, used experimental technologies such as carbon nanotubes and RRAM to create a prototype of a 3D chip that combines current computers’ CPU and RAM much faster in processing data and at the same time with a much lower power consumption.
Despite the steady progress and the resulting increase in speed, communications between current CPUs and RAM not to mention storage memory remain bottlenecks. As if that wasn’t enough, in recent years there’s a growing concern about the limits for the improvements still possible of silicon-based processors. Graphene seems to have become the equivalent of the philosopher’s stone but the transition from the creation of lab prototypes to commercial applications seems to be taking a long time.
In this scenario, there are those who are working on carbon-based solutions instead of silicon-based and the Stanford and MIT team also went that way. Graphene is the basis of this solution because 2D sheets of this material are used to create the nanocylinders that make up carbon nanotubes. They’re used to create the equivalent of CPUs but also RRAMs (resistive random-access memory), a new generation of memories.
Various technologies are also being developed to offer a new generation of memories that have the performance of current RAM but are non-volatile like current storage memories. This would overcome one of the biggest problems of current computers with an architecture that would provide a single type of extremely fast memory which at the same time would not require continuous rewriting, thus combining the best of the memories currently in use.
The Stanford and MIT team went beyond combining CPUs and RRAMs into a single chip with a 3D structure, a further progress compared to silicon chips. The problem is that producing silicon transistors requires temperatures of over 1,000° Celsius with the consequence that the production of a second layer above the first would damage it. Carbon nanotubes can be produced at temperatures below 200° C and the production of a second layer above the first wouldn’t damage it.
All this offers a number of advantages beyond the simple possibility to produce chips with a 3D structure that integrates CPU and RRAM. Carbon nanotubes require an amount of energy of a magnitude lower than silicon transistors and this also means that they produce less heat. For example, the prototype produced by the Stanford and MIT team contains 2 million transistors and 1 million RRAM cells.
For now, this is a prototype and we’ll need to see what kind of developments there will be. In May, HPE presented a prototype of The Machine, its computer of the future, which the company is developing for commercial purposes. In recent years, there were many announcements but commercial applications so far are very limited.
Max Shulaker and his team seem optimistic about the possibilities of their 3D chip based on carbon nanotubes, we’ll see in the coming years whether something’s coming out or it’s just one of the many experiments. However, it seems that what can be done with silicon can be done better with carbon but sophisticated technologies are required to use it in electronics.