Western Sydney University has revealed plans to deploy an FPGA-based supercomputer named “DeepSouth” to support its neurological research at the International Centre for Neuromorphic Systems (ICNS). The university claims that DeepSouth is the first facility capable of simulating networks “at the scale of the human brain.” The supercomputer, expected to be operational by April 2024, mimics biological processes in hardware and can emulate large networks of spiking neurons at a rate of 228 trillion synaptic operations per second, approaching the estimated rate of operations in the human brain.
ICNS Director Professor André van Schaik stated that the FPGA-based design is more efficient and consumes less power than traditional high-performance computing (HPC) simulations using graphics processing units (GPUs) and multicore central processing units (CPUs). The supercomputer is designed to accelerate the understanding of brain functions and facilitate brain-scale computing applications across various fields, including sensing, biomedical research, robotics, space exploration, and large-scale AI applications.
DeepSouth’s smaller size compared to most supercomputers is highlighted, and its hardware-reprogrammable nature, enabled by FPGAs, allows for the addition of new neuron models, connectivity schemes, and learning rules. Researchers can program the system in Python, eliminating the need to delve into the intricacies of the underlying hardware. The collaborative effort behind DeepSouth involved ICNS, the universities of Sydney and Melbourne, and Germany’s University of Aachen. Commercial partners supplying DeepSouth’s hardware have yet to be disclosed by Western Sydney University.
