Siggraph

Nvidia Makes Re-Lighting Quick and Easy

Changing lights in an animated scene–moving a key light to heighten drama, or adding a diffuse background glow to separate out a background from a foreground–can be painfully slow. For a fully built scene, a graphics processor might have to run through hundreds of thousands of calculations for each tweak of a lighting angle.
Occlusions are another graphics intensive roadblock: in a busy graphics production pipeline, occlusion culling (removing hidden surfaces before shading and rasterizing take place) is key for cutting down rendering time.

At Siggraph, Nvidia previewed a combination of hardware and software technology that promises to drastically reduce the time necessary for these tedious re-rendering operations, boosting interactivity.

“I can now redo the ambient occlusion for an entire (complexly shaded) scene in about 9 seconds, where previously it took 11 minutes,” says Eric Enderton, principal engineer on the Gelato renderer team.

Although the results are just shy of full, final film-level render quality, it was hard for me to see much of a difference. Enderton spotted the combo of the new Version 2.2 Gelato and G80 series GPUs (released earlier this year for the top end 5200/5600 series Quadro boards).

Don‘t forget your Mocha (so far just a codename for this subset of Gelato)–this is a greatly improved re-lighting pipeline that again brings speed improvements into the hundreds.

Other points of note in the upcoming Gelato release: faster raytraced reflections, more accurate subsurface scattering, enhanced texture conversion, freshened user interface, better data exporting functionality, and improved support for the Autodesk Maya Hair feature. For a cool user video (The Plush Life) that shows what Gelato Pro 2.2 can do (check out the new hair and velvet shader technology), swing on over to http://www.nvidia.com/gelatozone.

Also new at the show: Quadro Plex Visual Computing System (VCS) Model S4 1U graphics server, with four Quadro FX 5600 GPUs packed into a 1U housing. Off point perhaps, but Nvidia also launched the Tesla series cards at the show; these pull out some of the high-end visual chops of the Quadro cards, but, when combined with Nividia‘s Cuda coding initiative, allow HPC (high performance computing) types (read supercomputers) to code for heavy-duty parallel processing chores.