Intel Releases New AI Chip to Boost AI Development and Deployment Performance

2019-11-13

Intel showcased its new neural network processor (NNP) at the company's AI Summit in San Francisco on November 12, 2019. This is Intel's first dedicated ASIC chip designed to accelerate AI system development and deployment from the cloud to the edge, while providing complex and efficient deep learning services for cloud and data center customers. Intel also unveiled the next-generation Intel Movidius™ Myriad™ Vision Processing Unit (VPU) for edge media and computer vision inference applications.


Naveen Rao, Vice President and General Manager of Intel's Artificial Intelligence Products Group, said, "In the next phase of our AI plans, we will reach a breakthrough in computing hardware and memory. Dedicated hardware such as Intel Nervana NNP and Movidius Myriad VPU are essential for advancing AI. Using more advanced forms of system-level AI will help us transform data into information and information into knowledge." These products further enhance Intel's AI solution portfolio, which is expected to generate over $3.5 billion in revenue in 2019. Intel's AI product portfolio is the most comprehensive in the industry, enabling customers to develop and deploy AI models on devices ranging from large-scale cloud to tiny edge devices.


The new Intel Nervana NNP is now in production and delivered to customers. It is part of a system-level AI approach that provides a complete software stack, integrating open components and deep learning framework for maximum utilization.


Intel Nervana NNP-T achieves true balance between computing, communication, and storage, enabling near-linear high-energy efficiency scalability from small clusters to the largest Pod supercomputers. Intel Nervana NNP-I offers high energy efficiency and budget advantages, making it ideal for running intensive multi-modal inference in real-world scenarios. Both products are developed to meet the processing needs of leading AI customers such as Baidu and Facebook. Misha Smelyanskiy, Director of Facebook AI System Co-design, said, "We are excited to collaborate with Intel to deploy faster and more efficient inference computing models using Intel Nervana neural network processors, and extend support for our latest deep learning compiler, Glow, to NNP-I."


In addition, the next-generation Intel Movidius VPU, planned for release in the first half of 2020, combines unique and efficient architectural advantages, with inference performance expected to be over 10 times that of the previous generation and six times more efficient than competing processors. Intel also announced its new edge-focused Intel® DevCloud, which, together with the Intel® OpenVINO™ toolkit distribution, addresses a key challenge faced by developers, allowing them to prototype and test AI solutions on various Intel processors. With AI running on Intel® Xeon®™ Scalable processors in most regions worldwide, Intel continues to enhance the platform's AI inference performance across the entire Intel® processor lineup through Intel® Deep Learning Boost with Vector Neural Network Instructions (VNNI). While it will continue to serve as a powerful AI foundation for years to come, Intel's customers demand a doubling of performance every 3.5 months for their most advanced deep learning training requirements, and only a comprehensive AI solution portfolio like Intel's can achieve these breakthroughs. Intel has a holistic understanding of computing, memory, storage, interconnect, packaging, and software to maximize efficiency, programmability, and ensure the critical capability of scaling deep learning across thousands of nodes, thereby better serving customers.

RECENT AI NEWS

RECENT AI TOOLS