Datacenter Trends

Nvidia Touts Its 'Accelerated Computing' Vision for AI

The GPU maker has unveiled the CUDA-X AI, its latest technology to advance artificial intelligence in the datacenter.

• By John K. Waters
• 03/20/2019

Nvidia's 10th annual GPU Technology Conference, underway this week in San Jose, Calif., kicked off with a flashy video underscoring the company's artificial intelligence (AI) strategy, and CEO Jensen Huang fairly gushed about his company's new CUDA-X AI data science acceleration software.

Server acceleration, he said, is becoming a standard to boost AI in the datacenter.

"Accelerated computing is not just about the chips," Huang told a packed house at the San Jose State University Event Center. "Accelerated computing is a collaboration, a codesign, a continuous optimization between the architecture of the chip, the systems, the algorithm and the application."

Santa Clara, Calif.-based Nvidia has grown in a few short years into a multibillion-dollar company by marketing its graphical processing units (GPUs) to datacenter operators as the right silicon for processing the flood of data demanded by a new generation of AI-oriented applications.

Coming on the heels of the company's recent acquisition of high-performance networking company Mellanox, this new collection of software-acceleration libraries holds great promise for that market.

"We're seeing a great growth in data," said Mellanox CEO Eyal Waldman, who joined Huang on stage. "An exponential growth. And we're also starting to see that the old program-centric datacenter is changing into a data-centric datacenter, which basically means the data will flow and create the programming. Instead of us creating a program using the data, the data will start creating the program using the data itself. These are things that we can work out and actually get very synergistic architecture solutions for the future for the datacenter."

CUDA-X AI is built on top of Nvidia's CUDA parallel computing platform and programming model for general computing on GPUs. It's designed to provide essential optimizations for deep learning, machine learning and high-performance computing (HPC).

The software "unlocks the flexibility of our Nvidia Tensor Core GPUs," the company said in a blog post, to address the "end-to-end AI pipeline"  (data processing, feature determination, training, verification and deployment).

The CUDA-X AI acceleration libraries have been adopted by a number of companies, including PayPal, SAS, Walmart and Microsoft. Nvidia claims they can speed up machine learning and data science workloads by up to 50x. The software is integrated into the Nvidia T4 GPU servers offered by several top vendors, Huang said, including Cisco, Dell EMC, Fujitsu, HPE, Inspur, Lenovo and Sugon.

Huang also announced a new deep learning partnership with Amazon.com, which will use Nvidia's T4 Tensor Core GPUs for Amazon Web Services (AWS).

"We're going to need a large, large number of partners to be able to take these architectures that are really complicated -- to bring it out into the enterprise," Huang said.