Intel's First AI-Dedicated ASIC, Nirvana NNP
NNP-T for AI training and NNP-I for inference configuration
Next-generation Movidius VPU, 10 times faster than previous generation


To transform data into information, and information into knowledge, more advanced system-level AI is required.

On the 13th, Intel unveiled the 'Intel Nirvana NNP' for AI training and inference and the next-generation 'Intel Movidius Myriad VPU' at the AI Summit 2019 held in San Francisco, USA.

Intel's first application-specific integrated circuit (ASIC) product dedicated to AI, the Intel Nervana Neural Network Processor (NNP), consists of the NNP-T1000 for AI training and the NNP-I1000 for AI inference.

NNP-T provides energy-efficient scalability, and NNP-I is suitable for implementing multimodal inference. Both products were developed to meet the AI processing requirements of AI customers such as Baidu and Facebook.

The next-generation Intel Movidius Myriad Vision Processing Unit (VPU) is designed for edge media, computer vision, and inference applications. Intel predicts that the next-generation Intel Movidius Myriad VPU, set to be released in the first half of 2020, will be able to provide more than 10 times higher inference performance than the current Intel Movidius Myriad VPU 2.

Intel also announced the Intel OpenVINO Toolkit Distribution and Intel DevCloud for Edge. This allows developers to test and prototype AI solutions based on Intel processors before purchasing hardware.

Naveen Rao, Intel Vice President and General Manager of the AI Products Group, stated, “As AI continues to advance, we are reaching the limits of computing hardware and memory,” adding, “To sustain the advancement of AI, dedicated hardware such as the Intel Nirvana NNP or the Movidius Myriad VPU is necessary.”