Circuits hold the key to future computing

Editorial / August 13, 2024

Working on the frontiers of technology is not all about discoveries, it is also finding solutions to old bottlenecks to find fresh perspectives. 

Working on one such crucial aspect to mitigate circuit ageing, Dr. Sonal Singhal, Associate Professor, School of Engineering explores the possibilities of reinventing the von Neumann computing architecture. 

“Circuits designed at advanced technology node are prone to device degradation. Bias Temperature Instability (BTI) is the most common device degradation mechanism in Static Random Access Memory (SRAM). At advanced technology nodes, besides time-zero process variability, time-dependent process variability is a dominant degradation mechanism. Ensuring the lifetime reliability of advanced computing systems such as SRAM-based In-Memory Computing (IMC) is of paramount importance,” says Dr. Singhal.

Enhancing the age of a circuit leads to a path that navigates to a multitude of variables that include prolonging the life of a circuit, speed of processing, and energy efficiency to name a few. A crucial aspect of this research is to recognise how traditional computing is reaching its threshold and provide circuits that can make a computer work like a human brain. 

Neuromorphic computing, as it is called, besides requiring new materials to make chips also needs a fresh perspective on how circuits are made. 

These new computing technologies need multiple chips and technologies with billions of transistors all packed in an ultra-small form all working seamlessly. 

Advances in the actual architecture of neuromorphic systems point to a direction where neurons on chips can ‘learn’ as they compute. While software neural networks have been able to do this for a while, advancements in physical neuromorphic circuits are now showing results. This provides a new future in computing where the possibility of higher speed and more complexity at a lesser energy cost. 

At the university’s Very Large Scale Integration (VLSI) Design Laboratory, the Cadence Design Systems, Xilinx Vivado System Edition, FPGAs, Tanner Tools, Model Sim, Coventorware Turbo Extended Analyzer, High-Performance Computing Workstation and High-performance Computing Rack-Server are extensively used to find solutions.

Keeping in line with the university’s vision of collaborative research, this work is conducted with the University of Michigan, Shanghai Jiao Tong University, Joint Institute Shanghai, and BITS Pilani. These collaborations hold pivotal importance as they involve shared resources. 

“The VLSI Design Laboratory provides a fertile ground for Ph.D. scholars to engage. One of our scholars is currently developing an in-memory computing design with a focus on ensuring reliability. Previously, a Ph.D. scholar created a spiking neural network that prioritises reliability. He is currently working as a senior ASIC design engineer at Cyient Semiconductors,” says Dr. Sonal.

The work done also aligns with the UN Sustainable Development (SDG) Goal 9, which aims to build resilient infrastructure, promote sustainable industrialisation, and foster innovation; and SDG 7, which aims to double the global rate of improvement in energy efficiency.

With the world's annual energy consumption estimated to be 580 million terajoules, the research undertaken by Dr. Sonal also plays a crucial role in creating self-aware circuits that reduce the amount of power used. 

As we live on the cusp of history where digital technology may prove the starting point of something completely new. Just like a human brain that works both analog and digital at the same time, the computers of tomorrow may have circuits that act just like the human brain does.