The rapid evolution of Artificial Intelligence (AI) and its expansion into edge and end devices have spurred research into highly energy-efficient, low-latency computing paradigms, such as In-Memory Computing (IMC). The NeuroSoC project emerged to address these advancements. As part of the HORIZON EUROPE initiative, the NeuroSoC project brings together a diverse consortium of 12 partners from 9 countries. Their shared mission is to develop the next generation of edge AI processing hardware, tackling the challenges of deploying AI in resource-constrained environments where power consumption and cost are critical factors.
Addressing Edge AI Challenges
To meet the stringent power and real-time requirements of edge applications, the NeuroSoC project is exploring innovative analog technologies, rather than traditional digital approaches. This shift aims to achieve performance improvements that significantly surpass current solutions, addressing the growing demands on smart device applications.
NeuroSoC Objectives
A key goal of the NeuroSoC project is the development and demonstration of a state-of-the-art Multiprocessor System-on-Chip (SoC) that incorporates advanced Phase Change Memory (PCM) technology. This ambitious effort targets a remarkable 100-fold improvement in computational efficiency, making AI capabilities more accessible at the edge in a cost-effective and energy-efficient manner.
Ubotica’s Role in NeuroSoC
Among the leading partners in the NeuroSoC consortium is Ubotica, a recognised authority in edge-AI acceleration. Ubotica’s expertise is pivotal to the project’s success, as it leads a keywork package focused on developing a performance evaluation framework for the NeuroSoC accelerator. This framework includes establishing a comprehensive set of industry benchmarks to assess the performance of prevalent Deep Neural Network architectures across both established and emerging edge processors.
By providing these benchmarks, Ubotica helps to measure and optimise the NeuroSoC chip’s capabilities, ensuring that it meets the highest standards of performance and efficiency.
Looking Ahead
The NeuroSoC project represents a significant advancement in edge AI processing, combining cutting-edge technology with the collective expertise of its partners. Ubotica’s involvement underscores its commitment to driving innovation and advancing the frontiers of edge-AI acceleration. Together, they are set to redefine AI deployment across various sectors, marking a transformative step forward in the evolution of edge computing.