Mathematical algorithms mimicking brain neural networks built on computer chips
Process workloads up to 1,000 times faster than existing processes

Intel has revealed the status of its latest neuromorphic research system, Pohoiki Springs, which provides 100 million neuron computing power.
Pohoiki Springs is a rack-mounted system for data centers and is the largest neuromorphic computing system developed by Intel to date. It is a system that embeds 769 Loihi neuromorphic chips inside a physical frame (chassis) the size of five standard servers.

Inspired by the human brain, the Loihi processor can handle demanding workloads that operate like the human brain, not only up to 1,000 times faster than existing processors but also with up to 10,000 times greater efficiency.

Poiki Springs is a system that extends the Loihi architecture to verify its potential to solve a wide range of difficult computational tasks, including artificial intelligence-related problems.

Intel's smallest neuromorphic system, Kapoho Bay, consists of two Loihi chips with 262,000 neurons and supports real-time edge workloads.
Pohoiki Springs, with 100 million neurons, is an important step in increasing Loihi's neural capacity to the level of a small mammalian brain, enabling it to process larger and more sophisticated neuromorphic workloads.

Intel's neuromorphic system is similar to Pohoiki Springs and is currently in the research phase, providing tools to develop and characterize neural network-inspired algorithms.

Traditional general-purpose processors such as CPUs and GPUs are particularly adept at tasks that are difficult for humans to perform, such as precise mathematical calculations, but the role and application of recent technologies are expanding to become more sophisticated.

From automation to artificial intelligence, computers are like humansThis demands the emergence of an architecture that adapts to change while processing more active, unstructured, and noisy data in real time.

Neuromorphic computing fundamentally redesigns computer architecture. It applies the latest insights from neuroscience to create chips that are more similar to the human brain. Neuromorphic systems implement the organization, communication, and learning capabilities of neurons at the hardware level.

Mike Davis, Director of Intel’s Neuromorphic Computing Lab, stated, “Pohoiki Springs can scale the Loihi neuromorphic research chip by more than 750 times while consuming less than 500 watts of power,” adding, “Research partners can use this system to process workloads that run slowly on existing architectures, including high-performance computing (HPC) systems, faster.”