It seems that Moore's Law is not slow enough to improve the transistor density and cost, and the cost of designing chips and the cost of factories that etch chips are also rising. In order to keep IT innovation moving forward, any of these cost savings will be very welcome.
At present, one of the promising research areas in the chip design world is to use machine learning techniques to actually help deal with some tasks in the design process. The use of machine learning in chip design is also one of the topics that Jeff Dean, a senior researcher in Google ’s research department, talked during his keynote speech at the 2020 International Solid State Circuits Conference in San Francisco. He helped Google, a hyperscale provider, invented Many key technologies.
Facts have proved that Google is not a whim for the computing engine. Google is one of the world's largest consumers of CPUs and GPUs, and the designer of TPUs. TPUs range from the edge to the data center. They can be used for both machine learning inference and machine learning training. Therefore, this is not just the research activity of this search engine giant and public cloud provider-if it intends to continue to advance the TPU roadmap, if it decides to start designing its own custom Arm server chip, or decide to Designing customized Arm chips for mobile phones and other consumer-level devices is even more practical.