Intel Increases the Neuromorphic Research System to 100 Million Neurons


Intel has announced the latest and most powerful neuromorphic research system, located in Pohoiki Springs, Intel has scaled up their Neuromorphic Research System to feature 100 million neurons. Pohoiki Springs is a data center rack-mounted system, and this system integrated 768 Loihi neuromorphic research chips inside a chassis that is significantly larger than standard server chassis. Pohoiki Springs is considerably larger than Intel's previous neuromorphic system, Kapoho Bay, which features two Loihi chips with 262,000 neurons.

Intel has increased its neuromorphic research system from having two Loihi chips to having a system, Pohoiki Springs, that features 768 Loihi integrated neuromorphic research chips

Neuromorphic chips are very different when compared to Intel's other CPUs, and the neuromorphic chips are designed to mimic the human brain, this is done by basing the chips on neurobiology. Instead of impulses, which are the electrical shock sent to the brain, these chips have their variation called "spikes." While these chips are designed to mimic the human brain allows them to handle unprocessed or "noisy" data in real-time easily. At the same time, for conventional processors, this would be very time-consuming.

Intel To Disclose New Information on 14th Gen Meteor Lake & 15th Gen Arrow Lake CPUs at HotChips 34

Pohoiki Springs, a data center rack-mounted system unveiled in March 2020, is Intel’s largest neuromorphic computing system developed to date. It integrates 768 Loihi neuromorphic research chips inside a chassis the size of five standard servers. (Credit: Intel Corporation)

These systems are primarily used in data centers, and these data centers usually cater to vary of differences like Constraint satisfaction and Optimization problems. These problems are present in everyday life, in tasks as simple as package delivery planning to problems as complicated as allocating a stock portfolio to minimize risk at a target rate of return.

Since Intel has scaled up its neuromorphic system, from just two Loihi chips to having 768 integrated Loihi chips allows their system to handle much more complex problems at a much more efficient rate. Alongside that increase in processing power, this new system uses just 500 watts of power, which is very efficient for the 768 Loihi processing chips.

This system could be used in deep learning AI to not only process information faster but also to compute various pieces of information that humans may need immediately. While this upgrade to Intel's neuromorphic system adds 768 Loihi chips, for context, the human brain has the equivalent to roughly 650,000 Loihi chips. While this new system will not be competing against the human brain in terms of raw power at this point, this substantial upgrade shows how in the future, these neuromorphic systems may be more widespread than they currently are.