Intel Reveals World’s Largest ‘Brain-Inspired’ Neuromorphic Computer

In a recent development, scientists at Intel have built the world’s largest neuromorphic computer. The computer is designed and structured to mimic the human brain. Moreover, it is tipped to support future artificial intelligence (AI) research. The neuromorphic computer, dubbed “Hala Point,” can perform AI workloads 50 times faster. Furthermore, it can use 100 times less energy than conventional computing systems that use CPUs and GPUs. According to Intel representatives, these figures are based on findings uploaded on March 18 to the preprint server IEEE Explore, which have not been peer-reviewed.

Hala Point: A Neuromorphic Computer To Mimic Human Brain

Hala Point will reportedly be deployed at Sandia National Laboratories in New Mexico. Scientists will use it to solve problems in device physics, computing architecture, and computer science. It is powered by 1,152 of Intel’s new Loihi 2 processors and a neuromorphic research chip. It is pertinent to mention that this large-scale system comprises 1.15 billion artificial neurons and 128 billion artificial synapses distributed over 140,544 processing cores.

The brain-inspired device can make 20 quadrillion operations per second. It’s just amazing! Neuromorphic computers are capable of processing data differently from supercomputers. That’s why, it’s hard to compare them. Trinity is the 38th most powerful supercomputer in the world. It boasts approximately 20 petaflops of power. Moreover, the world’s most powerful supercomputer Frontier, boasts a performance of 1.2 exaFLOPS, or 1,194 petaFLOPS. Reports indicate that Hala Point performed a high energy efficiency reading for AI workloads of 15 trillion operations per watt (TOPS/W). Most traditional neural processing units (NPUs) and other AI systems execute satisfactorily under 10 TOPS/W.

Hala Point is a research prototype anticipated to feed into future commercial systems. These future neuromorphic computers might lead to large language models (LLMs) like ChatGPT learning constantly from new data. They are expected to reduce the massive training burden ingrained in existing AI deployments.