Reshaping 3D electronics for AI hardware
A new advancement in computer chip technology has emerged from collaborative research led by Sang-Hoon Bae, an assistant professor at the McKelvey School of Engineering, Washington University in St. Louis.
In a significant leap forward, Bae and his team, comprising international researchers from institutions like MIT, Yonsei University, Inha University, Georgia Institute of Technology, and the University of Notre Dame, have successfully demonstrated monolithic 3D integration of layered 2D materials in novel processing hardware for artificial intelligence (AI) computing.
Multifunctional computer chips, known for their integrated sensors, processors, memory, and specialised components, have traditionally expanded laterally, leading to increased time for information transfer between components. Bae likens this to constructing a house: “You build out laterally and up vertically to get more function, more room to do more specialised activities, but then you have to spend more time moving or communicating between rooms.”
The team's monolithic 3D-integrated chip signifies a paradigm shift from conventional laterally integrated computer chips. Comprising six atomically thin 2D layers, each fulfilling a distinct function, this device marks a substantial decrease in processing time, power consumption, latency, and physical footprint. Achieved through densely packed processing layers ensuring robust interlayer connectivity, this hardware showcases unparalleled efficiency and performance in AI computing tasks.
This breakthrough offers a new approach to integrate electronics, potentially heralding a new era in multifunctional computing hardware. Central to this technology is ultimate parallelism, which could vastly enhance the capabilities of AI systems, allowing them to perform complex tasks with unprecedented speed and accuracy. Bae elaborates: “Monolithic 3D integration has the potential to reshape the entire electronics and computing industry by enabling the development of more compact, powerful, and energy-efficient devices. Atomically thin 2D materials are ideal for this, and we are dedicated to refining this material until we can integrate all functional layers on a single chip.”
These devices are not only more compact and powerful but also offer increased flexibility and functionality, making them suitable for a broad range of applications. Bae envisions a wide spectrum of uses, from autonomous vehicles to medical diagnostics and data centres. One particularly promising application is in-sensor computing, which amalgamates sensor and computer functions into one device. This integration allows for obtaining a signal and directly computing data, resulting in faster processing, reduced energy consumption, and enhanced security, as data transfer is minimised.