Ai Contribution to Dramatic Increases in Computing Power
The development of Artificial Intelligence is in urgent need of core hardware upgrades. Traditional chip products cannot meet the demand for intensive linear algebraic computation and high throughput of massive data in terms of basic capabilities, and the development of artificial intelligence chips has naturally become a catalyst for the development of artificial intelligence. In general, the demand for deep neural networks for computing chips mainly involves two aspects of communication and computation. The communication aspect has to address the need for communication of massive amounts of data between the computing chip and the storage device, thus requiring both storage of massive amounts of data and a large cache, as well as high bandwidth that enables massive data interaction between the computing unit and the storage. The computational side has to address the large number of computational requirements needed to support dense linear algebraic computations (e.g., convolution, f...