Intel and Sandia National Labs Unveil Revolutionary $1.15B “Hala Point” Neuromorphic Research System

Neuromorphic computing, a field that continues to be explored, has seen significant advancements in recent years, particularly in the development of specialty chips. Today, Intel and Sandia National Laboratories are celebrating the deployment of the Hala Point neuromorphic system, which they claim to be the world’s highest capacity system. With a total of 1.15 billion neurons, Hala Point represents the largest deployment of Intel’s Loihi 2 neuromorphic chip, which was introduced in late 2021.

The Hala Point system consists of 1152 Loihi 2 processors, each capable of simulating a million neurons. Despite their small size (31 mm2 per chip with 2.3 billion transistors), these chips are built on the Intel 4 process, making them highly efficient. The entire system occupies just 6 rack units of space and consumes 2.6 kW of power. By surpassing the SpiNNaker system, Hala Point now holds the title for the largest disclosed neuromorphic system, albeit with only a slightly larger number of neurons compared to the British system, which requires 100 kW.

Hala Point replaces Sandia’s previous neuromorphic system, Pohoiki Springs, which was based on Intel’s first-generation Loihi chips. The new system offers ten times the number of neurons and significantly improved performance. Both systems are aimed at advancing Sandia’s research into neuromorphic computing, a computing paradigm that aims to mimic the brain’s behavior. The goal is to solve problems that traditional processors cannot handle efficiently.

Sandia plans to use Hala Point for large-scale neuromorphic computing research, surpassing the capabilities of Pohoiki Springs. With a simulated neuron count comparable to that of an owl brain, Sandia believes that a larger-scale system will enable them to fully exploit the potential of neuromorphic computing in solving real-world problems across various fields.

One area of interest for both Sandia and Intel is the application of neuromorphic computing in AI inference. As neural networks in AI systems attempt to emulate the human brain, there is a natural synergy with brain-mimicking neuromorphic chips, despite some algorithmic differences. Intel’s research on Hala Point has shown promising energy efficiency, reaching efficiencies as high as 15 TOPS-per-Watt at 8-bit precision. Additionally, neuromorphic systems eliminate the need for extensive data processing and batching, which is typically required for traditional processors.

One particularly intriguing use case is the potential to augment neural networks with additional data on the fly, without the need for costly re-training. This approach aligns with how brains continuously learn and augment datasets. However, this concept remains primarily within the realm of academic study for now.

Intel and Sandia envision systems like Hala Point leading to the development of commercial neuromorphic systems in the future, potentially at even larger scales. However, further research is needed to refine algorithms and optimize workload mapping for this style of computing to demonstrate its utility on a larger scale.

Background Information

---

About Intel: Intel Corporation, a global technology leader, is for its semiconductor innovations that power computing and communication devices worldwide. As a pioneer in microprocessor technology, Intel has left an indelible mark on the evolution of computing with its processors that drive everything from PCs to data centers and beyond. With a history of advancements, Intel's relentless pursuit of innovation continues to shape the digital landscape, offering solutions that empower businesses and individuals to achieve new levels of productivity and connectivity.