Google builds the largest machine learning center with the fourth-generation TPU accelerator

After announcing the fourth-generation TPU accelerator last year, at Google I/O 2022, Google announced that 8 sets of Cloud TPU v4 Pod computing devices will form the largest machine learning center, and emphasize the use of 90% carbon-free energy to drive.

In the previous description, Google said that the computing performance of the fourth-generation TPU accelerator has doubled compared to the previous generation, at the same time, the POD computing device composed of 4096 sets of fourth-generation TPU accelerators can also correspond to more than 1 exaFLOPS of computing performance. This time, the machine learning center is composed of 8 sets of Cloud TPU v4 Pod computing devices. The peak operation can reach up to 9 million teraflops per second, making it the world’s largest public machine learning hub, and will be open to the public through Google Cloud services.

Industry teams including Cohere, LG AI Research, Meta AI, and Salesforce Research have actually experienced Google’s machine learning center, and it can set up an interactive development environment through the TPU VM architecture, and can flexibly apply machine learning frameworks such as JAX, PyTorch, or TensorFlow. In addition, through fast interconnection and optimized software stack characteristics, it creates more outstanding performance and scalability.

The PaLM (Pathways Language Model) language model mentioned by Google this time at Google I/O 2022 is trained through 2 sets of TPU v4 Pods, so as to correspond to faster language translation and understanding.