by Heather Mackey

There’s a natural tendency at GTC to focus on 3D visuals, but 3D audio is also emerging as a topic here.

Sound engineers who work to produce 3D audio – commonly thought of as “surround sound” – can leverage GPUs for the processing required to simulate sound in actual spaces, with implications for gaming, home entertainment systems and film.

In a talk for audio professionals, Nicolas Tsingos, of Dolby Laboratories, described the inroads GPUs are making in 3D audio rendering and simulations. His presentation, “Interactive 3D Audio Rendering Systems” covered the ways that GPUs can be used in audio processing and in creating simulations of the way acoustics sound in three-dimensional environments. 

Starting with an overview of GPUs and parallelism, Tsingos showed how he and other acoustical researchers have applied many-core processors to acoustical simulation challenges, and then shared optimization secrets they’ve learned along the way. They have found that, depending on the scenario, GPUs can accelerate processing speeds by 2x to 300x versus using CPUs on their own.

The key to achieving faster performance is in writing to the GPU’s strengths, he noted. Because GPUs are well suited to processing large chunks of data, Tsingos advised session members that they would see better results by increasing data per transfer and reducing the number of transfers.

Attendees showed keen interest when Tsingos showed movies of various acoustical simulations. One displayed differences in sound, according to changes in an area’s contours. Another showed acoustical modeling of an echo from a Mayan temple in Mexico.

Several dozen audience members stayed behind for extended Q&A, which ranged from acoustical design and the “computing stereo of the future” to microphone array processing in real time.