Nvidia unveils technologies for deep learning supercomputers, autonomous cars and virtual reality

Nvidia announces its plans for supercomputers based on neural networks, autonomous cars and virtual reality


Nvidia announced a slew of new products and technologies in the 2016 GPU Technology Conference. The Tesla P100 GPU is a hyperscale data center accelerator. The GPU is meant for data centers, scientific and technical research, or churning statistics. Eight of these Tesla P100 GPUs in a box makes the Nvidia DGX-1 pictured above, the world's first deep learning supercomputer.

The DGX-1 is made for artificial intelligence applications, for tackling complex problems that cannot be handled by conventional programming. The supercomputer comes with the DIGITS platform (Deep Learning GPU Training System). This interactive platform can be used for designing deep neural networks. Demos showed by Nvidia showcased the supercomputer processing large volumes of audio and visual information. The Nvidia DGX-1 will be available in the US from June. One of the first uses of the DGX-1 is going to be at the Massachusetts General Hospital's Clinical Data Science Center. The AI capabilities and machine learning will be used for radiology, pathology and genomics. Radiology and Pathology in particular stand to benefit by advances in machine processing of visual data. The DGX-1 is also aimed at cloud service providers for processing images, classifying images and speech recognition.

Nvidia's other big announcement was the Iray VR, based on their Iray ray-tracing engine. Aimed at professionals and enterprise users, the virtual reality ray tracing engine will allow architects, lighting designers and product designers to showcase their creations in photo-realistic VR. Iray is designed to render materials, surfaces and textures more realistically. One of the challenges overcome was rendering in real time, which was handled by a number of light probes that are sampled based on where the user is looking.

A 3DS max plugin called Iray VR lite is also available, which could be used for mobile applications. The engine is platform and device agnostic, shown to be working on both the HTC Vive and Google Cardboard at the event. Nvidia also debuted Mars 2030, an interactive VR experience made with NASA, that lets users explore Mars. An outer space environment, a drilling exercise, a habitat and driving on a rover over Martian terrain are part of the VR experience.

Nvidia's final big announcement were new technologies for driverless cars. A custom mapping software uses the Drive PX 2, Nvidia's autonomous car driving platform in the car, and the new Tesla GPU in a data center to provide high quality maps. These detailed maps along with AI allow cars to understand and navigate their environment better. A futuristic announcement was the roborace championship, a racing event for driverless cars.  The cars will all be identical vehicles, designed by Daniel Simon, who conceptualized the Light Cycles for Tron: Legacy. All the cars will be equipped with Drive PX 2, and the only difference between two teams will be the software used to run the vehicles.

 

Find our entire collection of stories, in-depth analysis, live updates, videos & more on Chandrayaan 2 Moon Mission on our dedicated #Chandrayaan2TheMoon domain.