Google claims its 4th-gen TPU-based supercomputers, used for AI training, are up to 1.7x faster and 1.9x more power-efficient than Nvidia's A100 systems
Alphabet Inc's (GOOGL.O) Google on Tuesday released new details about the supercomputers it uses to train its artificial intelligence models …
3/ The big advance here is in scaling - keeping performance as you add thousands of chips to the system. The key to it is another custom Google technology, this time their optical switching technology.
5/5 Google split PaLM (540bln parameters) across two supercomputers for 50 days. The researchers say their flexible switching system allows for routing around hardware failures, etc. Again, link to the full paper: https://arxiv.org/...