On 6th December 2018, Applied Brain Research Inc. (Canada), a neuromorphic computing software developing company, released the results of a study that evaluated the performance of their Nengo Deep Learning Toolkit, running an audio keyword spotting deep learning network on Intel’s Loihi neuromorphic research chip and compared the energy efficiency to traditional hardware.
The benchmarking results show that Nengo DL on Intel Loihi uses 38x less energy per inference than an architecturally identical network running on an NVIDIA Quadro K4000 GPU. The study also compared the dynamic energy cost per inference performance of the same deep network on several other platforms. In each case, the Nengo DL on Loihi network consumed significantly less power. In comparison, the NVIDIA Jetson TX1 edge GPU consumed 7.3x more energy, the Intel Xeon E5-2630 CPU consumed 8.2x more energy, and the Movidius Neural Compute Stick consumed 1.9x more energy.
According to the Point Of View of Sachin Garg – AVP : Semiconductor and Electronics, at MarketsandMarkets™, The benchmark result indicates that Nengo DL on Loihi outperforms other platforms on an energy cost per inference basis while maintaining near-equivalent inference accuracy, which indicates that neuromorphic computing will have a major role to play in Artificial Intelligence. This development is an important breakthrough towards the commercialization of neuromorphics. Currently, AI relies on GPU acceleration for training and inference. However, the latest study suggests that neuromorphic computing, once commercialized, could be more efficient for real-time AI processing than GPU and CPU platforms.
Neuromorphic Computing Market:
Neuromorphic computing involves the use of the functional principles of a human brain to help design and fabricate an artificial system. Neuromorphic circuits are inspired by the nervous system and are useful components in artificial perception/action systems, which also help in verifying neuro-physiological models. The major drivers for the growth of the neuromorphic computing market include new ways of computation possible due to the end of Moore’s law; requirement of better performing ICs for computation speed, power consumption, packaging density; and increase in demand for artificial intelligence and machine learning.
The neuromorphic computing market is expected to be valued at USD 272.92 million by 2022, growing at a CAGR of 86.0% between 2016 and 2022.
Impact on Artificial Intelligence Market:
Artificial Intelligence is gaining momentum owing to its agile structure and offers a high level of satisfaction and customer retention. Moreover, AI combines data from multiple resources and uses it as a knowledge hub, which ultimately results in accurate predictions about consumer needs.
The Artificial Intelligence Market was valued at USD 16.06 billion in 2017 and is expected to reach USD 190.61 billion by 2025, growing at a CAGR of 36.62% during 2018–2025. AI is an evolving market and needs systems with powerful compute capabilities to process and analyze large quantities of data. Compute-intensive chipsets are a critical parameter in processing AI algorithms; the faster the chipset, the faster it can process data required to create an AI system. Increasing technological advancements are expected to enable smaller, more efficient, and powerful neuromorphic chip-based systems that replace large hardware devices in the coming years.
In April 2015, Intel stated that it could keep Moore’s law going for 10 more years by developing 7nm and 5nm fabrication technologies. This indicates that beyond 2023 or 2025, the size of an IC cannot be shrunk, even by doubling transistor count, as this would result in reduced space between electrons and holes and lead to problems such as current leakage and overheating in ICs. These problems would lead to slower performance, high power consumption by ICs, and reduced durability. Thus, the need to find an alternate method to increase the computational power of chips has fueled the development of neuromorphic chips.
The increasing need for collection, analysis, and decision-making from complex and unstructured data is driving the demand for better performing processors that may outpace both, CPU and GPU architectures. Companies such as ABR and Intel have been working on specialized architectures and high-performance computing in order to keep pace with the evolution of technology.