Future trend and exploration of AI chip

Apr 25,2021



1. Neuromorphic Chip


Neuromorphic chips are chips that subvert the classic von Neumann computing architecture and use electronic technology to 

simulate the proven workings of biological brains to build chips that resemble biological brains.


Advantages of neuromorphic chips:


(1) computing and storage fusion, breaking through the bottleneck of Memory Wall.

(2) decentralized multi-core architecture with powerful fine-grained interconnectivity.

(3) better online learning. Tsinghua University, Intel, IBM and other schools and enterprises are doing research in this area.


 Future trend and exploration of AI chip


2. Reconfigurable Computing Chips


Reconfigurable computing chips, also known as software definition chips, address the following problems and mission requirements of current AI chips:


(1) there is a difficult balance between efficiency and flexibility.

(2) complex AI tasks require the combination of different types of AI algorithm tasks.

(3) different tasks require different computational accuracy. The design idea of reconfigurable computing chip is that the hardware and software can be programmed, 

     and the hardware architecture and function can change with the change of the software.


Newsletter

CONTACT US

Contact us