Google's machine learning specialized processor 'TPU v5e' with cost performance increased by 2.3 times is now available on Google Cloud
Google's cloud computing service 'Google Cloud' now allows processing using the machine learning specialized processor '
Cloud TPU v5e is generally available | Google Cloud Blog
https://cloud.google.com/blog/products/compute/announcing-cloud-tpu-v5e-in-ga/?hl=en
TPU v5e is the 5th generation model of Google's proprietary machine learning specialized processor ' Tensor Processing Unit (TPU) ', and at the time of its announcement, it had a performance per dollar of 2.5 compared to the previous generation model TPU v4. The appeal was that it is twice as high and the inference latency is 1.7 times lower.
Google announces the 5th generation model of AI specialized processor TPU 'TPU v5e', up to 2 times the training performance and up to 2.5 times the inference performance per dollar compared to the previous model - GIGAZINE
Google announced on November 9, 2023 that it has made TPU v5e available on Google Cloud. At the same time, we also released performance measurement results for TPU v5e.
The graph below shows the cost performance of machine learning using TPU. TPU v5e exhibits cost performance more than 1.5 times that of TPU v4 in bfloat16 (bf16) , and when performing int8 quantization, cost performance improves by 2.3 times.
Google also announced that it will strengthen its collaboration with Anthropic , an AI company that develops ChatAI rival Claude . Anthropic has already introduced TPU v5e and is currently researching and developing the language model that will form the basis of Claude.
Google Announces Expansion of AI Partnership with Anthropic
https://www.prnewswire.com/news-releases/google-announces-expansion-of-ai-partnership-with-anthropic-301981815.html
Related Posts:
in Hardware, Posted by log1o_hf