This morning, Intel announced the general readiness of Pohoiki Springs, a powerful self-contained neuromorphic system that’s about the size of five standard servers. It says the system will be available to members of the Intel Neuromorphic Research Community via the cloud using Intel’s Nx SDK and community-contributed software components, giving them a tool to scale up their neuromorphic research and explore ways to accelerate workloads that run slowly on today’s conventional architectures.
With 768 Loihi chips and 100 million neurons spread across 24 Arria10 FPGA Nahuku expansion boards (which contain 32 chips each) that operate at a power level of under 500 watts, Intel claims Pohoiki Springs is similar in neural capacity to the brain of a small mammal — an ostensible step on the path to supporting larger and more sophisticated neuromorphic workloads. Just this week, Intel demonstrated that the chips can be used to “teach” an AI model to distinguish among 10 different scents.
“Pohoiki Springs enables our research partners to explore ways to accelerate workloads that run slowly today on conventional architectures, including high-performance computing systems,” said Intel neuromorphic compute lab director Mike Davies in a statement.
Above: Pohoiki Springs, a datacenter rack-mounted system unveiled in March 2020, is Intel’s largest neuromorphic computing system developed to date.
Neuromorphic engineering, also known as neuromorphic computing, describes the use of circuits that mimic the nervous system’s neuro-biological architectures. Researchers at Intel, IBM, HP, MIT, Purdue, Stanford, and others hope to leverage it to develop a supercomputer a thousand times more powerful than any today.
Intel cautions that neuromorphic systems like Pohoiki Springs are still in the research phase and are not intended to replace conventional computing systems. But it points out that they provide a tool for researchers to develop and characterize new neuro-inspired algorithms for real-time processing, problem-solving, adaptation, and learning.
For instance, preliminary research suggests that neuromorphic chips like Loihi excel at constraint satisfaction problems, which require evaluating a large number of potential solutions to identify the one or few that satisfy specific constraints. Neuromorphic systems have also been shown to rapidly identify the shortest paths in graphs and perform approximate image searches, as well as to mathematically optimize specific objectives over time in real-world optimization problems.
Loihi architecture
Intel’s 14-nanometer Loihi chip has a 60-millimeter die size and contains over 2 billion transistors, 130,000 artificial neurons, and 130 million synapses, as well as three managing Lakemont cores for orchestration. Uniquely, Loihi features a programmable microcode engine for on-chip training of asynchronous spiking neural networks (SNNs), or AI models that incorporate time into their operating model such that the components of the model don’t process input data simultaneously. Intel claims this will be used for the implementation of adaptive self-modifying, event-driven, and fine-grained parallel computations “with high efficiency.”
According to Intel, Loihi processes information up to 1,000 times faster and 10,000 more efficiently than traditional processors, and it can solve certain types of optimization problems with more than three orders of magnitude gains in speed and energy efficiency. Moreover, Loihi maintains real-time performance results and uses only 30% more power when scaled up 50 times (whereas traditional hardware uses 500% more power), and it consumes roughly 100 times less energy than widely used CPU-run simultaneous location and mapping methods.
Above: A closer look shows one of the rows within Intel’s latest neuromorphic research system, Pohoiki Springs.
The Loihi development toolchain comprises the Loihi Python API, a compiler, and a set of runtime libraries for building and executing SNNs on Loihi. It provides a way to create a graph of neurons and synapses with custom configurations, such as decay time, synaptic weight, and spiking thresholds, and a means of simulating those graphs by injecting external spikes through custom learning rules.
The debut of Pohoiki Springs follows that of Intel’s smallest neuromorphic system, Kapoho Bay, which comprises two Loihi chips with 262,000 neurons and supports a variety of real-time edge workloads. Intel and INRC researchers have used it to recognize gestures in real time, read braille using novel artificial skin, orient direction using learned visual landmarks, and more while consuming only tens of milliwatts of power.
In July, Intel took the wraps off of Pohoiki Beach, a 64-chip computer capable of simulating 8 million neurons in total. Like Pohoiki Springs, it was made available to research partners like Accenture to “advance the field” and scale up AI algorithms such as spare coding and path planning.
Author: Kyle Wiggers.
Source: Venturebeat
