Deep learning at the hardware level Intel will push artificial intelligence chips next year

Recently, at the IDF16 (Intel Developer Forum) conference held in San Francisco, the United States, Intel told developers that next year will continue to develop its Xeon Phi series processors, and launched special products for the field of artificial intelligence.

Unlike traditional CPUs, the Xeon Phi series was first officially launched by Intel in November 2012 as a coprocessor with strong scientific computing capabilities. Traditionally, although the traditional single-processor computing power is very powerful, it is costly. It is not worthwhile for some tasks such as scientific computing, where the amount of computation is very large but can be split into many small steps for parallel processing. Based on this basic idea, the earliest Xeon Phi internally designed 60 cores that can be calculated in parallel, specifically designed to handle some of the large and complex mathematical problems that can be resolved.

It is because of this feature that Xeon Phi plays a major role in the system as a co-processing task for a graphics card GPU. The CPU will dump the branch tasks that require a huge amount of calculation and time-consuming to coprocessors such as Xeon Phi, thus ensuring the efficient operation of the entire system.

Currently, although Xeon Phi has been widely deployed in various data analysis areas and is also used in almost all areas of deep learning, most customers will also deploy artificial intelligence coprocessors from Nvidia to share them together. The computational burden of the CPU.

Based on this status, Intel said that in the future, it will increase the Xeon Phi's ability to serve as a host processor, change the traditional architecture of the CPU's add-on processing, and allow users to implement huge computing by simply using Xeon Phi's single chip plus memory.

At the same time, Intel said that it will internally incorporate a deep learning algorithm at the chip level to implement the hardware, making the new Xeon Phi play a greater role in the frontier areas of extensive use of deep learning such as voice processing, image recognition, and autopilot.

In fact, due to the decline of the PC market and the inability of the mobile market, Intel has long regarded the field of artificial intelligence as its future development direction. In addition to vigorously developing artificial intelligence chips, Intel also acquired startup company Nervana Systems last week, which focuses on the development of deep learning chips and software.