Intel and China’s Baidu Collaborates to Develop Neural Network Training Chip
Baidu is hosting its Create conference for AI developers at Beijing, and it has announced a new partnership with Intel at the same. The two companies have plans to work together to build a new Nervana Neural Network Processor for training, aka NNP-T 100.
The Nervana Neural Network Processor for training is a 16nm NNP-T, and as the name suggests, it is specifically designed for the purpose of deep learning and research of the neural network. The two companies have collaborated for both the software and hardware manufacturing of the chip, such that it will provide full support for the Baidu’s PaddlePaddle deep learning framework. The PaddlePaddle framework focuses on deep learning and research of the neural networks, to utilize it for the AI applications. Intel already provides its support for the framework through its Intel Xeon Scalable processors.
The chip is being built under the code name Spring Crest and is based on a new class AI model. According to Intel, the chip has got a different architecture than the other chips made by Intel. The new NNP-T 100 is enabled with image recognition and has got the software support for the on-chip memory. The company claims that the chip is much faster at performance and has been equipped with the high-speed on- and off-chip interconnects.
The corporate vice president of Intel Naveen Rao made the announcement during the event, where he said, “The next few years will see an explosion in the complexity of AI models and the need for massive deep learning compute at scale. Intel and Baidu are focusing their decade-long collaboration on building radical new hardware, codesigned with enabling software, that will evolve with this new reality – something we call ‘AI 2.0.’”
The company had released the first version of NNP in 2017 and plans to ship this version of NNP by the end of this year. The two companies have been working together since 2016, and their future collaborative plans also include MesaTEE, a memory-safe function-as-a-service (FaaS) computing framework based on the Intel Software Guard Extensions (SGX) technology.
Yashica is a Software Engineer turned Content Writer, who loves to write on social causes and expertise in writing technical stuff. She loves to watch movies and explore new places. She believes that you need to live once before you die. So experimenting with her life and career choices, she is trying to live her life to the fullest.