Google has revealed details about its artificial intelligence supercomputer, which is faster and more efficient than Nvidia systems. Google has been designing and deploying AI chips called Tensor Processing Units (TPUs) since 2016, and the new supercomputer is a testament to their success. With over 90% of the market for AI model training and deployment, Nvidia still dominates the industry, but Google’s new supercomputer shows that they are a serious contender in the AI space. A new supercomputer is a powerful tool for machine learning models, and it is sure to have a major impact on the tech industry.
Google is a major AI pioneer, having developed some of the most important advancements in the field over the last decade. However, some believe the company has fallen behind in terms of commercializing its inventions. To prove it hasn’t squandered its lead, Google has been racing to release products and services, a “code red” situation internally. AI models and products such as Google’s Bard or OpenAI’s ChatGPT require a lot of computers and hundreds or thousands of chips to work together to train models, with the computers running around the clock for weeks or months. (wbctx.com) Google is now working hard to make sure its AI products and services are commercially viable, and that its lead in the field is not lost.
Google has developed a new system with over 4,000 TPUs (Tensor Processing Units) and custom components to run and train AI models. This system, called TPU v4, is 1.2x-1.7x faster and uses 1.3x-1.9x less power than the Nvidia A100. It was used to train Google’s PaLM model, which competes with OpenAI’s GPT model, over 50 days. TPU v4 supercomputers are the workhorses of large language models due to their performance, scalability, and availability. Google’s new system is a major breakthrough in the field of AI and will help to further advance the development of AI models.
Google’s TPU chip has been compared to other AI chips in an industry-wide test called MLperf, and the results show that it is a powerful and efficient chip. However, the latest Nvidia AI chip, the H100, was not included in the comparison as it is more recent and was made with more advanced manufacturing technology. According to Nvidia CEO Jensen Huang, the results for the H100 were significantly faster than the previous generation, with 4x more performance than the A100. This shows that the H100 is a powerful chip for training large language models with great energy efficiency.
AI requires a lot of computer power, which can be expensive. To reduce the cost, companies are developing new chips, components, and software techniques. Cloud providers such as Google, Microsoft, and Amazon are taking advantage of the power requirements of AI by renting out computer processing and providing credits or computing time to startups. Google’s TPU chips are being used by Midjourney, an AI image generator, as an example of how cloud providers are helping to reduce the cost of AI.
