Baidu announced that it has developed its own chip for artificial intelligence. Called Kunlun, the new chip was built to accommodate "high performance requirements of a wide variety of AI scenarios", the company said at its Baidu Create conference.
The new product range includes training chip '818-300' and inference chip '818-100'. Kunlun can be applied to both cloud and edge scenarios, such as data centres, public clouds and autonomous vehicles, according to the Baidu. It leverages Baidu’s AI ecosystem, which includes AI scenarios like search ranking and deep learning frameworks like PaddlePaddle.
In 2011, Baidu started developing an FPGA-based AI accelerator for deep learning and began using GPUs in data centres. Kunlun, which is made up of thousands of small cores, has a computational capability which is nearly 30 times faster than the original FPGA-based accelerator. Other key specifications include 14nm Samsung engineering, 512 GB/second memory bandwidth, as well as 260TOPS while consuming 100 Watts of power.
In addition to supporting the common open source deep learning algorithms, Kunlun chip can also support a wide variety of AI applications, including voice recognition, search ranking, natural language processing, autonomous driving and large-scale recommendations, the company said. Baidu plans to continue to develop upon this chip, to enable the expansion of an open AI ecosystem. This includes a focus on applications for intelligent vehicles, intelligent devices, voice recognition and image recognition.
