Nvidia Shows Off Its AI, Deep Learning, VR Capabilities at GTC

Highlights from Nvidia’s GPU Technology Conference range from technology aimed at making VR more real to products for accelerating innovation in deep learning.

Attendees take a break outside the San Jose Convention Center, and next to one of two “Drone Zones” at the show, where companies showed off their Nvidia-powered wares.

During his keynote, Huang introduced a number of new offerings designed to help developers and systems makers build deep learning capabilities into their products and accelerate innovation around AI. Among the new products was the Tesla P100, a massive chip based on Nvidia’s new 16-nanometer Pascal architecture that packs 150 billion transistors (from GPU cores and memory components) and uses a 16nm FinFET manufacturing process for better performance and efficiency, High-Bandwidth Memory (HBM2) technology and Nvidia’s NVLink interconnect.

The DGX-1 combines eight Tesla P100 GPUs with two Intel Xeon server chips to deliver up to 170 teraflops of performance in a 3U (5.25-inch) form factor. It will be priced at $129,000 and initially will be delivered to AI research institutions.

On the show floor, a number of companies showed off AI capabilities powered by Nvidia GPU technologies. One of those vendors, DeepArt, uses a deep learning algorithm that enables the computer to turn photos into pieces of art that can be painted in a range of styles chosen by the user.

Nvidia announced its new Iray VR rendering technology, which uses such capabilities as ray tracing to create virtual worlds that Huang said are “photoreal.” “Realistic is not enough,” the CEO told attendees. “It’s got to be real. It’s got to be photoreal.” He used the technology to give the audience a virtual tour of Nvidia’s upcoming new headquarters.

Apple co-founder Wozniak beamed into the convention center during Huang’s keynote and used Nvidia’s technology to take a virtual tour of the Red Planet.

At the VR Village on the show floor, attendees could test out VR headsets from Oculus and HTC and play games running on Nvidia technology.

Huang introduced the Drive PX 2, a board that comes with two Tegra chips and two Pascal discrete GPUs. The development kit, which carmakers can use to develop their autonomous cars, pulls in data from sensors on the vehicle to get a picture of the environment and software then decides how to react to the data. The development board will ship later this year.

Roboraces will pit driverless cars from 10 teams competing against each other. Each team gets two cars, and all cars will be equipped with the Drive PX 2 technology. It will be up to the teams to develop the algorithms and software for the cars. The races will be part of the new Formula E ePrix series.

A BMW SUV on the exhibit floor showed how Nvidia’s Drive technology, armed with 3D sensing software from Quanergy Systems, can help vehicles detect pedestrians and automatically hit the brakes before they hit the pedestrians.

An autonomous, three-wheeled cart developed at the Massachusetts Institute of Technology is designed to move without a driver.

The upcoming Wave Glider SV3 platform from Liquid Robotics is powered by wave and solar energy and uses Nvidia GPUs to help drive its autonomous capabilities. Due out later this year, the boat can run on its own for as long as a year.

Drones are a fixture at GTC, with two “Drone Zones” set up—one inside the convention center, the other outside. Here an employee with Intelligent Flying Machines shows off an Nvidia-powered indoor drone that can be used in large warehouses and factories.

Dell, Hewlett Packard Enterprise (HPE), Lenovo and Cisco Systems were among a large number of OEMs that showed off systems that take advantage of GPU accelerators from Nvidia. Pictured here is Dell workhorse PowerEdge R730, which recently was upgraded with Intel’s latest Xeon E5 chips and also uses GPU accelerators.

At the HPE booth, attendees were greeted by an ice sign that doubled as a dispenser for chilled drinks. In the booth, HPE showed off a range of systems, from workstations to HPC offerings.