Latin American engineers help build the world’s largest neuromorphic system (Hala Point) to enable more sustainable AI

Intel announced that it has built the world’s largest neuromorphic system, in which Engineers from Intel’s Guadalajara Design Center in Mexico participated (GDC for its acronym in English).

This neuromorphic system called Hala Point, use the Intel Loihi 2 processor and aims to boost research for future brain-inspired artificial intelligence (AI), addressing challenges related to the efficiency and sustainability of current AI. Hala Point is the second generation of large-scale neuromorphic computing systems, with architectural improvements that achieve more than 10 times the neural capacity and up to 12 times the performance compared to its predecessor, Pohoiki Springs.

Hala Point is the first neuromorphic system at scale demonstrating next-generation computational efficiencies in conventional AI workloads. Characterization shows that it can support up to 20 quadrillion operations per second, or 20 petaops, with efficiency exceeding 15 trillion 8-bit operations per second per watt (TOPS/W) when running deep neural networks. This exceeds the levels achieved by architectures based on graphics processing units (GPUs) and central processing units (CPUs). Hala Point’s unique capabilities could in the future enable real-time continuous learning for AI applications such as scientific and engineering problem solving, smart city infrastructure management, logistics, large language models ( LLM) and AI agents.

The system is a 6U rack-mountable chassis, which inside contains 36 Hala Point N3x cards with 32 Loihi 2 chips each, for a total of 1,152 chips. This results in a system of 1.15 billion neurons, which at maximum capacity can run 20 times faster than a human brain and up to 200 times faster at lower capacity. While Hala Point is not intended for neuroscience modeling, its neural capacity is roughly equivalent to that of an owl’s brain or a capuchin monkey’s cortex.

At GDC, the functional and physical design of the card, mechanical design, heatsinks and thermal simulations were carried out in order to meet the product specifications.

Since 2018, Intel Labs – Intel’s research unit – at GDC has been responsible for the design and development of the neuromorphic computing hardware platforms and systems of the “Neuromorphic Compute Lab” (NCL), and was a key player in the development and on-time delivery of Hala Point at Sandia National Laboratories. At GDC, the functional and physical design of the card, mechanical design, heatsinks and thermal simulations were carried out in order to meet the product specifications. Hala Point was assembled and tested at Intel Guadalajara, in addition, researchers from Intel Labs Guadalajara devised part of the algorithms that allowed the characterization of the system and achieved the aforementioned performance, all thanks to the close collaboration and support of various engineering groups in GDC.

Recent trends in scaling deep learning models with millions of parameters have exposed enormous sustainability challenges in AI and highlighted the need for innovation at all levels of hardware architecture. Neuromorphic computing is a fundamentally new approach that draws on insights from neuroscience that integrate memory and computation with highly granular parallelism to minimize data movement. In published results from the International Conference on Acoustics, Speech and Signal Processing (ICASSP) in April, Loihi 2 demonstrated orders of magnitude gains in the efficiency, speed and adaptability of emerging small-scale edge workloads.

Loihi 2 neuromorphic processors apply brain-inspired computing principles such as asynchronous and event-driven spiking neural networks (SNN), integrated memory and compute, and sparse and continuously changing connections to achieve orders of magnitude gains in power consumption. energy and performance. Neurons communicate directly with each other rather than through memory, reducing overall power consumption.

Loihi-based systems can perform AI inference and solve optimization problems using 100x less power at speeds up to 50x faster than conventional CPU and GPU architectures.. Although still under investigation, future neuromorphic LLMs capable of continuous learning could generate gigawatt-hours of energy savings by eliminating the need for periodic retraining with ever-growing data sets.

Currently, Hala Point is a prototype that will advance commercial system capabilities in the future. It is anticipated that these lessons will lead to practical advances, such as the ability of LLMs to continually learn from new data. Such advances will significantly reduce the training load of AI systems. Researchers at Sandia National Laboratories plan to use Hala Point for advanced computing research. The organization will focus on solving scientific computing problems in device physics, computer architecture, and computational science.

Together with an ecosystem of more than 200 members of the Intel Neuromorphic Research Community (INRC), the company is working to push the boundaries of brain-inspired AI and advance this technology from research prototypes to industry-leading products. industry in the coming years.

For Latest Updates Follow us on Google News


NEXT Unreleased enemy concepts with influences from Metal Gear and Titanfall » Hero Network