nVidia Moves Graphics Rendering to the Clouds

SAN FRANCISCO – nVidia has announced a platform for rendering and delivering highly-detailed 3D images over the Web to client devices, ranging from a smart phone to a PC, that takes the rendering load off the client.

The RealityServer platform for cloud computing is a combination of nVidia’s (NASDAQ: NVDA) Tesla processors (and soon Fermi, its new GPU family), rendering software and streaming software to send it to the client.

Rendering a richly-detailed image, such as the 3D images in Google Earth, puts the load on the client, noted Dan Vivoli, senior vice president of marketing at nVidia, at a media event Tuesday. “That’s fine so long as you don’t want to get too rich in fidelity with the scene. What you want is to get where that’s no longer a barrier, to get where you no longer worry about your computer for rendering the images.”

The RealityServer comes in three forms: Medium, Large and Extra Large, but nothing in a size small. The Medium server has eight GPUs and is designed for tens of concurrent users. It’s aimed at a local design team. The Large server has 32 GPUs and is meant for large production teams with hundreds of concurrent users. The XL has 100 GPUs and is a consumer service platform suitable for thousands of concurrent users.

The next ingredient is the rendering software from Mental Images, which does extremely highly detailed, precise photorealistic imagery in seconds instead of days. Mental Images’ iray (CQ) ray tracing software employs the massively parallel CUDA architecture of NVIDIA GPUs to create the images by properly simulating the physics of light in its interaction with objects.

“What we’re trying to do is enable 3D graphics on the Web and make 3D a ubiquitous experience on the Web,” said Vivoli. “All we send to the client is images. Video is a series of images generated on the RealityServer for each user.”

Vivoli added that nVidia is not planning to be in the business of providing cloud computing cycles to people, just selling the hardware, and that at this point, the chief impediment is rendering the frames as quickly as possible. The bandwidth requirements are no worse than they are for streaming video.

Jon Peddie, president of Jon Peddie Associates, said the RealityServer will be great for certain markets. “The idea of being able to view beautiful images easily from the Web has been discussed for some years. The problem has always been that you had to download a client and have sufficient graphics capability in your machine and had to have pretty high bandwidth. We have all that now,” he told InternetNews.com.

But its use, its consumption, will likely be confined to certain markets for now. It won’t, for example, be in Google Earth any time soon, he said.

“While all that stuff is fascinating, it’s interesting, it’s hard to say immediately what the business model is beyond the conventional business model today, like architects and home buyers and whatnot. It’s revolutionary in concept, but the consumption of it will be evolutionary,” said Peddie.

The nVidia RealityServer platform will be available November 30, 2009. A developer edition of RealityServer 3.0 software will be downloadable free of charge, including the right to deploy non-commercial applications November 30, 2009.

Get the Free Newsletter!

Subscribe to our newsletter.

Subscribe to Daily Tech Insider for top news, trends & analysis

News Around the Web