Ari Dharma researcher arrogantly introduced, CPU, GPU as a general-purpose computing chip, designed to deal with thread logic and graphics, high power consumption when dealing with AI computing problems, low cost, in the field of AI computing need dedicated architecture chip to solve the above problems.
At present, Alibaba's independently developed AI chips are mainly used to solve AI inference problems in commercial scenes such as images, video recognition, and cloud computing to improve computational efficiency and reduce costs.
According to arrogance, Ali-NPU developed independently by Alibaba, based on the advantages of a large number of algorithm models accumulated in the AI field by Alibaba's machine intelligence technology labs, etc., designing microstructures and instruction sets based on the AI algorithm model, and achieving the maximum amount at the minimum cost. AI model algorithm operation.
According to the design, Alibaba's Ali-NPU performance will be 10 times that of mainstream CPUs and GPU architecture AI chips currently on the market, while the manufacturing cost and power consumption are only half, and the cost-effectiveness is more than 40 times. In the future, Ali-NPU capabilities, Not only can we better meet the video and image processing needs, but we can also use Alibaba Cloud to output the computing power and empower all walks of life.
According to statistics, Ali has previously invested in the Cambrian, Barefoot Networks, Shen Jian, Kneron, ASR, Zhongtian Micro and many other chip companies.