Intel has recently announced the creation of Hala Point, the world’s largest neuromorphic system, marking a significant step towards more sustainable and efficient artificial intelligence. Deployed initially at Sandia National Laboratories, Hala Point uses Intel’s advanced Loihi 2 processor and builds on the success of its predecessor, Pohoiki Springs, by offering substantial improvements in architecture. This enhancement boosts neuron capacity by more than tenfold and performance by up to twelve times.
“The computing cost of today’s AI models is rising at unsustainable rates. The industry needs fundamentally new approaches capable of scaling. For that reason, we developed Hala Point, which combines deep learning efficiency with novel brain-inspired learning and optimization capabilities. We hope that research with Hala Point will advance the efficiency and adaptability of large-scale AI technology,” said Mike Davies, director of the Neuromorphic Computing Lab at Intel Labs.
Hala Point distinguishes itself by being the first large-scale neuromorphic system capable of demonstrating state-of-the-art computational efficiencies on mainstream AI workloads. It can support up to 20 quadrillion operations per second, or 20 petaops, and offers unprecedented efficiency exceeding 15 trillion 8-bit operations per second per watt (TOPS/W) when executing conventional deep neural networks.
Researchers at Sandia National Laboratories will use Hala Point for advanced brain-scale computing research, focusing on scientific computing problems across various domains such as device physics, computer architecture, and informatics. “Working with Hala Point improves our Sandia team’s capability to solve computational and scientific modeling problems. Conducting research with a system of this size will allow us to keep pace with AI’s evolution in fields ranging from commercial to defense to basic science,” stated Craig Vineyard, Hala Point team lead at Sandia National Laboratories.
While Hala Point remains a research prototype, Intel envisions its lessons will significantly enhance future commercial systems’ capabilities, notably enabling large language models to learn continuously from new data and reducing the training burden of AI deployments.
The drive for increasingly large deep learning models has exposed significant sustainability challenges within AI, necessitating innovation at the fundamental levels of hardware architecture. Neuromorphic computing, inspired by neuroscience, integrates memory and computing within a highly parallel framework to minimize data movement. This approach has demonstrated remarkable gains in efficiency, speed, and adaptability, as evidenced by Loihi 2’s performance at this month’s International Conference on Acoustics, Speech, and Signal Processing (ICASSP).
Hala Point integrates 1,152 Loihi 2 processors and supports up to 1.15 billion neurons and 128 billion synapses, distributed over 140,544 neuromorphic processing cores, within a six-rack-unit data center chassis. Its massively parallelized fabric offers significant memory bandwidth and communication speeds, providing a robust foundation for bio-inspired spiking neural network models.
Intel’s ongoing development of neuromorphic systems like Hala Point aims to address power and latency constraints that currently limit the real-world deployment of AI. With the continued collaboration of the Intel Neuromorphic Research Community (INRC), Intel is committed to advancing this brain-inspired technology from research prototypes to commercial products.
