Testing of Chips Used for Artificial Intelligence

After decades of promise, Artificial Intelligence (AI) or Machine Learning is finally at an inflection point. After being delegated to labs and selective applications, we are now seeing the beginning of the early stage of wide-spread adoption from medicine, and self-driving cars, to how factories are run, and to personal digital assistants such as Siri, Alexa and Google Assistant. The advent of power Graphical Processing Units (GPUs) and multi-core specialized computing chips have enabled the reality of commercially viable AI platforms. Many companies are developing their own version of high-performance AI chips, now also begin referred to as IPU or Intelligent Processing Units. These are high-performance chips that draw high current and consume a lot of power. Most of these are fabricated on the leading-edge semiconductor technology node. The advanced nodes are also typically low-voltage nodes and hence for a certain amount of wattage, the current consumption becomes much higher. Formfactor has been working with the leading-edge companies world-wide to develop probe cards that are highly optimized to test these IPU chips. This paper is written in collaboration between Keystone and Formfactor. The paper will provide an overview of the AI chip market and the key challenges to test AI chips. There will be a description of the solution to meet key testing requirements followed by analysis of data taken at customer site and a preview of future work.