Google Cloud to launch GPU offering in early 2017
Wed 16 Nov 2016
Google Cloud has announced its plans to add a series of GPU (Graphics Processing Unit) offerings to its service starting early next year.
According to a blog post, written by Google Cloud Platform product manager John Barrus, the company is offering the AMD FirePro S9300 x2 and two Nvidia Tesla options, the P100 and the K80, for deep learning, artificial intelligence and HPC applications.
Google is playing catch-up with rivals such as Amazon Web Services, Microsoft Azure and IBM’s Bluemix, which already provide GPU as a service. However, the tech giant hopes that its varied GPU offering and ‘affordable’ hourly pricing structure will win over developers.
Barrus explained that although CPU-based machines in the cloud are ideally suited for general purpose computing, more complex tasks such as rendering and large-scale simulations are best placed on more specialised GPU processors.
GPUs contain hundreds of times as many computational cores as CPUs and can significantly accelerate certain tasks, such as ‘risk analysis, studying molecular binding or optimizing the shape of a turbine blade,’ noted Barrus.
Google Cloud’s GPU service will be made available from early 2017 via the Google Compute Engine and the Google Cloud Machine Learning platform. Costings are yet to be revealed.
In a simultaneous announcement, Rob Craft, head of Google Cloud Machine Learning, posted that the unit has established a machine learning group specifically focused on the development of cloud-based machine learning services. The group will be headed by top machine intelligence researchers Fe-Fei Li and Jia Li.
Google also confirmed the general availability of its text analysis service, Cloud Natural Language API. Features include improved accuracy for entity recognition and ‘granular sentiment analysis at the sentence level.’ The tool will also support morphology and syntax analysis, which allows it to identify text features such as numbers, genders, person and tense.