Surprisingly late, Google Cloud will add GPUs (graphics processing units) as a service early next year, according to a blog post today. Amazon Web Services, Microsoft Azure and IBM’s Bluemix all already offer GPU as a service.
Google may be seeking to distinguish itself, however, with the variety of GPUs it’s offering. They include the AMD FirePro S9300 x2 and two offerings from NVIDIA Tesla: the P100 and the K80. And Google will charge by the minute, not by the hour, making GPU usage more affordable for customers needing it only for short periods.
CPU-based machines in the cloud are good for general-purpose computing, but certain tasks such as rendering or large-scale simulations are much faster on specialized processors, Google explained. GPUs contain hundreds of times as many computational cores as CPUs and excel at performing risk analysis, studying molecular binding or optimizing the shape of a turbine blade.
Google’s GPU services will be available in early 2017 through Google Compute Engine and Google Cloud Machine Learning.
Separately, Google said that it has created a machine-learning group focused exclusively on delivering cloud-based machine learning services. It will be led by researchers Fei-Fei Li, director of the AI lab at Stanford, and Jia Li, director of research at Snapchat. And Google released its natural-language API, a text-analysis service.
