On Tuesday at the Hot Chips conference in Palo Alto, California, Intel announced that it will launch two new artificial intelligence processors that will be part of the Nervana Neural Network Processor (NNP) product line to accelerate artificial intelligence. Among them, NNP-T is used for training and NNP-I is used for reasoning.
These chips were developed by a research and development center in Haifa, Israel. Intel has previously invested in some Israeli artificial intelligence startups, and the Nervana NNP series is named after its acquisition of Nervana Systems in 2016.
Nervana NNP-T, code-named Spring Crest, is used to build large-scale deep learning models from scratch. It can process data from a variety of deep learning models within the power budget while increasing memory efficiency. NNP-T can be directly integrated with popular deep learning frameworks such as Baidu’s PaddlePaddle, Facebook’s PyTorch, and TensorFlow.
On the other hand, the Nervana NNP-I, code-named Spring Hill, specializes in deep learning reasoning for major data centers. Intel said the processor is based on a 10-nanometer process and is based on the existing Ice Lake processor. By using a dedicated AI inference calculation engine, NNP-I can handle higher workloads with minimal energy consumption. It is also highly programmable without compromising performance or power efficiency.
With the increasing demand for complex computing in the field of artificial intelligence, Intel believes that the new artificial intelligence chip will help companies better handle the business of artificial intelligence.