Next-Generation Two-Head Computer Chips EPFL engineers have developed a computer chip that combines two functions — logical operations and data storage — into a single architecture, paving the way for more efficient devices. Their technology is particularly promising for applications based on artificial intelligence. Credit: EPFL / LANES 2020
EPFL engineers have developed a computer chip that combines two functions – logical operations and data storage – into a single architecture, paving the way for more efficient devices.
It is a great breakthrough in the field of electronics. Engineers at EPFL’s Nanoscale Electronics and Structure Laboratory (LANES) have developed a next-generation circuit that will enable smaller, faster and more energy-efficient devices – which would have great benefits for artificial intelligence systems. Their revolutionary technology is the first to use a 2D material for what is called a memory logic architecture, or it is the only architecture that combines logical operations with a memory function. The findings of the research team were published on November 4, 2020 Nature.
Until now, the energy efficiency of computer chips was limited by the von Neumann architecture currently in use, where data processing and data storage are performed in two separate units. This means that data must be transferred continuously between the two units, using considerable time and energy.
By combining the two units into a single structure, engineers can reduce these losses. That’s the idea of the new chip developed at EPFL, even if it takes a step beyond logical memory devices. The EPFL chip is made from MoS2, which is a 2D material consisting of a single layer with a thickness of three atoms. It is also an excellent semiconductor. LANES engineers have already studied the specific properties of MoS2 a few years ago, seeing that it is particularly suitable for electronic applications. Now the team has carried out this initial research to create a new generation technology.
The EPFL chip is based on moving gate effect transistors (FGFETs). The advantage of these transistors is that they can hold electric charges in the long run; they are commonly used in flash memory systems for cameras, smartphones, and computers. Unique electrical properties of MoS2 to be particularly sensitive to charges stored in FGFETs. This has enabled LANES engineers to develop circuits that function as a memory storage unit and as a programmable transistor. Using MoS2, many processing functions could be integrated into a single circuit and then changed as desired.
Deep knowledge
“This ability of circuits to perform two functions is similar to how the human brain works, where neurons are involved in storing memories and performing mental calculations,” says LANES chief Andras Kis. “Our circuit design has several advantages. Transferring data between memory units and processors can reduce the associated power loss, reduce the time required for computing operations, and reduce the space required. This opens the door to smaller, more powerful and more energy-efficient devices. ”
The LANES research team has also gained in-depth knowledge of manufacturing 2D materials circuits. “We made our first chip by hand ten years ago,” says Kis. “But since then we’ve developed an advanced manufacturing process that allows us to make 80 or more chips in a single turn, with well-controlled properties.”
Reference: Guilherme Migliato Marega, Yanfei Zhao, Ahmet Avsar, Zhenyu Wang, Mukesh Tripathi, Aleksandra Radenovic and Andras Kis, “Logic-in-memory based on atomic semiconductor in a based”, 4 November 2020. Nature.
DOI: 10.1038 / s41586-020-2861-0