Intel is unveiling the world's largest neuromorphic system with a neuron capacity on the level of an owl's brain, and expectations are high that it will help support AI research that mimics the future brain structure and solve challenges related to the efficiency and sustainability of AI.

Intel announced the world's largest neuromorphic system on the 18th.

This large-scale neuromorphic system, codenamed “Hala Point,” was first built at Sandia National Laboratories.

The goal is to support AI research that mimics the brain structure of the future using Intel's Loihi 2 processor and to solve challenges related to the efficiency and sustainability of current AI.

Hala Point is an evolution of Intel's first-generation large-scale research system, Pohoiki Springs, with architectural improvements that increase neuron capacity by more than 10x and improve performance by up to 12x.

Hala Point packages 1,152 Loihi 2 processors produced on Intel's 4-processor node into a 6-rack-unit data center chassis the size of a microwave oven.

This system supports up to 2,It supports up to 1.15 billion neurons and 128 billion synapses distributed across 145,444 neuromorphic processing cores that consume 600 watts of power.

The system also includes more than 2,300 embedded x86 processors for auxiliary operations.

Hala Point integrates processing, memory, and communication channels into a massively parallel fabric, delivering a total of 16 petabytes per second (PB/s) of memory bandwidth, 3.5 PB/s of core-to-core communication bandwidth, and 5 TB/s of chip-to-chip communication bandwidth. The system can process more than 380 trillion 8-bit synapses and 240 trillion neuron operations per second.

Applied to a biologically inspired spiking neural network model, the system can run 1.15 billion neurons at full capacity, 20 times faster than the human brain, and up to 200 times faster at lower capacities.

Although not intended for neuroscience modeling, the Hala point has a neuronal capacity roughly equivalent to that of an owl brain or the cortex of a capuchin monkey.

Loihi-based systems can perform AI inference and solve optimization problems up to 50 times faster and using 100 times less energy than conventional CPU and GPU architectures.

Hala Point is the first large-scale neuromorphic system to demonstrate state-of-the-art compute efficiency when applied to mainstream AI workloads.

The analysis results show that it can support up to 20 trillion operations per second, or 20 petaops, with an efficiency exceeding 15 TOPS/w 8-bit operation processing (TOPS is a unit operation that can be performed per second, and 1 TOPS is 1 billion unit operations performed per second) when running existing deep neural networks.

These figures match or surpass what GPU- and CPU-based architectures have achieved.

Hala Point’s unique capabilities can enable real-time continuous learning for AI applications such as scientific and engineering problem solving, logistics, smart city infrastructure management, large language models (LLMs), and AI agents.

Researchers at Sandia National Laboratories plan to use Hala Point for advanced brain-scale computing research. The institute will focus on solving scientific computing problems in device physics, computer architecture, computer science, and informatics.

“The compute costs of today’s AI models are increasing at an unsustainable rate. The industry needs a fundamentally new approach that can scale,” said Mike Davies, director of Intel Labs’ Neuromorphic Computing Lab. “That’s why Intel developed Hala Point, combining deep learning efficiency with novel brain-like learning and optimization capabilities. We hope that research into Hala Point will advance the efficiency and adaptability of large-scale AI technologies.”