PC gamers were left wanting after NVIDIA’s opening keynote at the GPU Technology 2016, developers were left positively giddy at the news of new technologies the firm will be releasing throughout the year.

Three of the new technologies that NVIDIA Chief Executive Officer, Jen-Hsun Huang showed off during the keynote are centred around VR development, deep learning and autonomous racing cars.

Let’s dive into these new products then shall we?

Iray VR

Built from NVIDIA’s own Iray ray-tracing engine, Iray VR is said to bring “breakthrough photoreal” visuals to VR.

The Iray VR plugin creates light probes which in turn renders lightfields. The lightfields reveal information about how light is traveling through a space.

This information is then cross referenced with where the viewer is standing and sends information about how the light should appear to the user.

“Iray VR is going to be unbelievable for people designing cars, for people architecting [sic] buildings and many other areas,” said Huang during the keynote.

During the conference Steve Wozniak joined Huang on stage via a live video feed and used an HTC Vive to explore the surface of Mars in a rover.

“Whooaaa, I feel like I’m actually here. I’m entering the rover,” Woz exclaimed in glee during the stream.

MARSEXPLORERS2
A virtual reality recreation of the surface of Mars.

There will also be a “lite” version of the software, Iray VR Lite which is not designed to render in 3D as well as Iray. That said developers can use Iray VR Lite for ray-tracing and creating VR experience for your smartphone and entry level VR headsets such as the Gear VR.

Both iterations of Iray will be available in June.

Drive PX 2

The Drive PX 2 is a super computer that will give vehicles artificial intelligence when working with Tesla based GPUs, such as the P100 (which we’ll get into in a bit), in data centres to navigate.

Driverless electric racing series, Roborace Championship will feature 10 teams each using two cars powered by the Drive PX 2 supercomputer.

Image by Chief Design Officer Daniel Simon / Roborace Ltd.
Image by Chief Design Officer Daniel Simon / Roborace Ltd.

Teams will have to program the AI in the vehicles to be ultra strategic to win and the more the cars race, the faster and smarter they will become.

The PX 2 is also capable of deciphering input from sensors such as a GPS, radar and cameras. While the idea of a life sized Scalextric track does excite us, point of the Roborace Championships is that the races will ultimately lead to safer and smarter autonomous cars.

DGX-1 Deep Learning System

NVIDIA calls the DGX-1 first ever deep learning super computer in a box and claims the box contains as much power as 250 servers.

Inside the box is the NVIDIA’s latest GPU, the Tesla P100 which is based off the Pascal architecture which NVIDIA has been talking up for two years.

The long awaited Pascal architecture from NVIDIA is here, though its only available in the DGX 1.
The long awaited Pascal architecture from NVIDIA is here, though its only available in the DGX-1.

Despite using 16nm transistor technology, NVIDIA has crammed 15 billion transistors onto the 600mm square chip. The result of this is that the Tesla P100 is capable of 5.3 TeraFLOPS of double precision single floating-point performance and 10.6 TeraFLOPS of single precision performance.

NVIDIA_DGX-1_KV_2000px

Over and above this NVLink will make communication between the CPU and GPU faster, this should allow data sharing rates to increase by five to 12 times compared to PCIe Gen3.

Orders for the DGX-1 are open but there is as of time of writing, there is no word on price.

[Source – NVIDIA] [Image – CC BY/2.0 Laineema]