Google’s self designed chip TPU offers better performance per watt than existing chips for machine …

Google has developed its own custom chip, which it has been using for more than a year now. The chip is known as Tensor Processing Unit, TPU, which has been named after the TensorFlow software that the search engine giant uses for its machine learning programs. Some even call the chip as an accelerator chip as it increases its speed in specific tasks. Explaining about the chip at Google’s I/O conference on Wednesday, CEO Sundar Pichai said that TPU provides better performance per watt than existing chip for machine learning tasks. Though it is not going to replace CPUs and GPUs, it can speed up machine learning process without consuming massive amount of energy. It has become essential to have such a chip as machine learning has become quite common.…


Link to Full Article: Google’s self designed chip TPU offers better performance per watt than existing chips for machine …