Published on Wednesday, 22 June 2011
NVIDIA PhysX 3.0 SDK has new modular architecture and a rewritten engine to improve
performance and the reliability of simulations across multi-core gaming devices and the
new tablets, smartphones and other handheld platforms.
|The PhysX SDK, built for realistic and real-time simulation based effects for games, film and other media, has been updated to version 3.0 to handle the larger game levels gamers are expecting, which require more actors. In PhysX 3.0, developers can combine multiple actors into a single ‘aggregate’, managed as one bounding-box entity in the broadphase stage of the collision pipeline, when the number of colliding pairs is estimated. This reduces the computing load required to predict collisions between actors, and should help to improve memory efficiency over earlier versions.|
|NVIDIA says PhysX 3.0 streams asset data into a simulation more efficiently through a new mechanism, binary in-place serialization, which allows memory-efficient insertion of actors into a scene. Also, out-of-scene actor creation, which allows actors to be created outside the scene and stored rather than being created and destroyed on demand, now manages assets better and helps control compute load spikes.
PhysX 3.0 works across multi-core devices including PCs, notebooks and consoles, as well as the recent handheld devices, tablets and smartphones. The new Task Manager has a managed thread pool allowing games to use multi-core processors on all platforms to increase performance and improve the gaming experience.
As well as an optimized physics runtime, NVIDIA is releasing tools for artists that have been tailored to work within the developer’s asset production pipeline. Developers will also have a new release of PhysX Visual Debugger for performance profiling, detailed memory analysis and improved visualization of PhysX content across major platforms.
PhysX 3.0 is available for PC, Xbox360, PLAYSTATION3, OSX, Linux and Android. PhysX 3.0 is designed to run on various CPU architectures. Performance can be accelerated by CUDA-enabled NVIDIA GPUs, including any GeForce 8-series or higher graphics card. www.nvidia.com/physx