IBM introduces AI processor AIU: using 5nm process, containing 23 billion transistors

Thanks to CTOnews.com netizen OC_Formula for the clue delivery! CTOnews.com, October 22, IBM Research recently released an AI processor called artificial Intelligence Unit (Artificial Intelligent Unit,AIU), which is IBM's first full SoC to run and train deep learning models. IBM claims that it works faster and more efficiently than GM CPU.

AIU is a prototype ASIC designed to support multiple formats and simplify artificial intelligence workflows from image recognition to natural language processing. The SoC contains 32 processing cores based on 5-nanometer technology, and the whole chip contains 23 billion transistors.

IBM says the AIU is as easy to use as a graphics card and can be plugged into any computer or server with a PCIe slot. IBM said it would soon share information about when AIU will be available.

CTOnews.com learned that AIU is from the AI hardware Center at the IBM Research Institute, which was launched in 2019 with the goal of increasing AI hardware efficiency by two and a half times a year. By 2029, IBM's goal is to train and run AI models 1000 times faster than in 2019. IBM says industrial-scale hardware will need to predict hurricanes or recessions, and AIU brings the goal a step closer.