Intel Aims for Efficient Brain-Scale Neuromorphic AI Research with the Loihi 2-Powered Hala Point
Equivalent in neuronal capacity to an owl's brain, Hala Point has been deployed at Sandia National Laboratories.
Intel has announced its development of what is claimed to be the world's largest neuromorphic computing system to date, deployed at Sandia National Laboratories for research into brain-inspired artificial intelligence (AI) systems and the sustainability of broad AI deployment: Hala Point, powered by the company's Loihi 2 processor.
"The computing cost of today's AI models is rising at unsustainable rates," Intel's Mike Davies, director of the neuromorphic computing lab at a company which has pledged to usher in the age of the "AI PC," admits.
"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."
Building on Intel's earlier Pohoiki Springs system, Hala Point is claimed to deliver ten times the neuronal capacity and up to 12 times the performance — translating into a compute performance of 20 peta-operations per second (POPS) and an energy efficiency of 15 tera-operations per second per watt (TOPS/W) for eight-bit precision in 10:1 sparse networks.
This, the company points out, "rivals and exceeds" the efficiency of competing designs that use general-purpose graphics processing units (GPGPUs) for acceleration, thanks to the brain-inspired neuromorphic Loihi 2 processor architecture.
"Working with Hala Point improves our Sandia team's capability to solve computational and scientific modeling problems," says Sandia National Laboratories' Craig Vineyard of the machine, which is to be used to advance research into brain-scale neuromorphic computing and how it can solve real-world computing problems in everything from computer architecture to device physics.
"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."
Hala Point is, Intel admits, a "research prototype" which will "advance the capabilities of future commercial systems" — though it is also working with Ericsson Research to use the same Loihi 2 processors to optimize the efficiency of cellular network infrastructure.
With 1,152 Loihi 2 processors and more than 2,300 x86 processors packed into a six-unit rack space, Hala Point is the densest deployment of Intel's neuromorphic technology yet — delivering 240 trillion neuronal operations every second and with a capacity positioned at the level of an owl brain or the cortex of a capuchin monkey.
More information on Loihi 2 is available in Intel's technology brief.