IT House May 14 news, according to tom’sHardware, at the I/O conference, Google released a preview of Google Cloud’s latest machine learning cluster, which is not only designed to achieve 9 exaflops peak performance, but also uses 90 % of carbon-free energy. It will be the largest public machine learning center in the world.
The heart of the new cluster is the TPU V4 Pod. These tensor processing units were announced at Google I/O last year, and AI teams from companies like Meta, LG, and Salesforce are already using the pods. The V4 TPU allows researchers to use the framework of their choice, whether it’s Tensorflow, JAX, or PyTorch, and has enabled Google Research to achieve breakthroughs in areas such as language understanding, computer vision, and speech recognition.
Based on Google’s Oklahoma data center, the cluster’s underlying workload is expected to be similar, munching on data from the fields of natural language processing, computer vision algorithms and recommender systems.
Access to clusters is provided in slices, from four chips (one TPU VM) all the way up to thousands.Slices with at least 64 chips utilize a three-dimensional ring link, providing higher bandwidth for collective communication operations. The V4 chip also has access to twice the memory of the previous generation — up from 16 to 32GiB — and doubles the speedup when training large models.
Google Research and Artificial Intelligence. “To make advanced AI hardware easier to use, we launched the TPU Research Cloud (TRC) program a few years ago, which provides free TPU access to thousands of ML enthusiasts around the world,” said Jeff Dean, senior vice president. Published hundreds of papers and open-source github repositories on topics ranging from “Writing Persian Poetry with Artificial Intelligence” to “Using Computer Vision and Behavioral Genetics to Distinguish Sleep and Exercise-Induced Fatigue.” The release of Cloud TPU v4 is great for Google Research and our This is an important milestone for the TRC program, and we are excited to collaborate on a long-term basis with machine learning developers around the world to use AI together.”
Google’s sustainability commitment means the company has been matching the energy use of its data centers with age-old energy purchases since 2017, and plans to dedicate its entire business to renewable energy by 2030. The V4 TPU is also more power efficient than previous generations, producing three times the FLOPS per watt than the V3 chip.
Access to Cloud TPU v4 Pods includes evaluation (on-demand), preemptive and committed use discount (CUD) options and is available to all Google AI Cloud users.
.
[related_posts_by_tax taxonomies=”post_tag”]
The post Google Officially Releases TPU V4 Chip appeared first on Gamingsym.