Since NVIDIA GPU parallel computing is suitable for deep learning of artificial intelligence, it has recently become a leader in AI chips, and it is completely overpowered by CPUs Intel and AMD, but with flexible field-programmable gate arrays (FPGA) chip performance, power consumption, and computing power increase. Buying heavily Altera, a major FPGA company, Intel, which fully started chip integration, and Xilinx, which is the latest acquisition target of Broadcom, are fearful. The road to NVIDIA's independence, coupled with Google, Amazon and Apple are actively engaged in chip development. Various types of ASIC chips are in full bloom. In 2018, AI chips are in full swing. AI applications are rapidly gaining popularity in various fields. Hundreds of billions of dollars in scale, will lead to a new wave of industrial innovation, to attract all the people competing to increase investment, holding AI chip computing technology is expected to preemptively obtain AI battlefield speaking rights, making AI chips warfare is fierce. From the current various ways People and horses are eager to participate in the development and integration of chip platforms. The AI wars are being played by chip makers. In recent years, NVIDIA has been highly sought after by the global market. As for Intel, AMD, Qualcomm, and Branch, IBM, Google, Apple, Facebook, Amazon, Microsoft (Microsoft) also accelerated the development of AI chips, all kinds of ASIC chip makers also rise up in an all-round way. Chip makers pointed out that with various AI chip computing power, performance and power consumption improved , can meet more Yuan AI applications, although NVIDIA grabbed the first battle victory of the AI chip, but faced with challenges from all walks of life since 2018, the advantage may no longer be. Chip makers said that AI killer application does not appear, it is still difficult to determine NVIDIA GPUs are the most suitable chips for deep learning. There are many existing algorithms. There are no chips that can meet the needs of all applications. NVIDIA GPUs are the first to get stuck, mainly relying on the advantages of parallel computing technology and the architecture designed specifically for AI, and vigorously develop each one. Deep learning software, libraries and tools to provide a more complete deep learning solution to accelerate deep neural network and training efficiency. Intel was previously suppressed by NVIDIA due to poor CPU implementation AI, and then quickly expanded by the merger strategy Increased AI strength. In 2015, it purchased Altera, an FPGA manufacturer, for 16.7 billion US dollars. It used FPGA computing power and flexibility to make up for CPU shortcomings. In 2016, Intel acquired AI and depth. Nervana Systems, a developer, bought another $ 15.3 billion in mobile tech maker Mobileye in 2017, and Intel AI softened the benefits of hardware technology integration. Intel's new Nervana AI platform has diversified advantages, Intel Xeon expandable products The series provides highly scalable computing capabilities for continuously evolving AI workloads, and a dedicated chip code-named Lake Crest for the most intensive deep learning training; Intel Mobileye is designed specifically for applications such as active secure autopilot; Intel FPGAs are programmable accelerators that perform deep learning inference; Intel Movidius low-power vision technology allows machine learning to be executed on a variety of end devices. FPGA maker Xilinx also consumes less energy than CPUs Performing AI operations faster, and recently gaining market support for deep learning neural network calculus, including Baidu and Amazon, have begun to use them to expand computing tasks that cannot be achieved by NVIDIA GPUs. Xilinx FPGA products have received support from many manufacturers. It also invests in deep knowledge of the mainland, strengthens deep compression, compiler tool chain and system level optimization machine learning In addition, as Broadcom merged with Qualcomm to break the market, Broadcom is optimistic about the application of FPGA technology, and may move to the acquisition of Xilinx. This also makes Xilinx, which is part of Bitcore China's ASIC miner supply chain, once again become a global focus. The chip makers believe that the future AI battlefield will be the killing of international companies, and Xilinth, currently alone, has a great chance of being incorporated.
