Google has announced the launch of its new Cloud TPU v5p, an updated version of its Cloud TPU v5e, boasting faster chips and improved performance, especially in training large language models like GPT3-175B. The v5p features 2x improvement in FLOPS and 3x improvement in high-bandwidth memory compared to the v5e, making it significantly faster and more cost-effective. Google’s DeepMind and Research teams have observed 2x speedups for large language model training workloads using TPU v5p chips compared to the performance on TPU v4 generation.