AI News from GTC 2019
March 21, 2019

AI News from GTC 2019

Artificial intelligence (AI) was on everyone’s lips at Nvidia’s GTC 2019, from the vendors and partners on the jam-packed exhibit floor, to announcements from Nvidia itself. In fact, the term “data science” – a branch of computer science involving large datasets – was prevalent at just about every twist and turn, as more and more people become involved in AI training and inference, machine learning, data analytics, and deep learning.

Among the major announcements in this area are:

CUDA-X AI SDK

Nvidia is putting the science and the data into the term “data science” with CUDA-X AI libraries. The libraries accelerate every complex step in a typical AI workflow.

CUDA-X AI is a collection of software acceleration libraries built on top of CUDA. These include cuDNN for accelerating deep learning primitives, cuML for machine learning algorithms, TensorRT for optimizing trained models for inference, and so on. These libraries work seamlessly with Nvidia Tensor Core GPUs to accelerate the development and deployment of AI-based applications. In essence, they enable an end-to-end platform – data processing, feature determination, training, verification, and deployment – for the acceleration of data science, speeding machine learning and data science workloads by up to 50x. 

CUDA-X AI libraries are available free as individual downloads or as containerized software stacks from the Nvidia NGC software hub. They can be deployed on desktops, workstations, servers, or on cloud computing platforms. 

Servers Optimized for Data Science

In another announcement, a number of large systems manufacturers – including Cisco, Dell EMC, Fujitsu, Hewlett Packard Enterprise (HPE), Inspur, Lenovo, and Surgon – have unveiled Nvidia-powered enterprise servers optimized for data science. The servers have been tuned to run Nvidia’s data science acceleration software.

The servers feature Nvidia T4 GPUs and draw only 70 watts of power; they are fine-tuned to run Nvidia CUDA-X AI acceleration libraries. 

In addition to the validated servers available today, several other partners have started the validation process for their T4 servers.

Amazon Web Services to Offer T4 GPUs 

Amazon Web Services announced that its new Amazon Elastic Compute Cloud (EC2)  instances featuring Nvidia T4 Tensor Core GPUs will be available in the next few weeks. The new G4 instances will provide AWS customers with a platform for efficiently deploying AI services. Through the AWS Marketplace, customers can pair the G4 instances with Nvidia GPU acceleration software, including the newly announced CUDA-X AI libraries for accelerating deep learning, machine learning, and data analytics.

The new EC2 G4 instances support the next generation of computer graphics for workflow acceleration.