AMD announces Open Physics Initiative
AMD bringing GPU physics to the masses
By Jason Cross | PC World | Published: 10:55, 01 October 2009
A press release from AMD today announces the development of an Open Physics Initiative, together with Pixelux Entertainment. Pixelux is the developer of the Digital Molecular Matter (DMM) physics system used first in LucasArts' Star Wars: The Force Unleashed.
The period of exclusivity with LucasArts is over, and now Pixelux will license its technology to others. As part of this agreement, Pixelux will integrate its proprietary technology with Bullet, a free and open-source physics engine (used in recent games like Trials HD and Free Realms).
According to the press release, AMD is actively pursuing the development of the Bullet physics library on OpenCL and DirectX 11's DirectCompute. On top of this, Pixelux is pursuing development of its finite element physics simulation software on OpenCL and integrating it with Bullet. Note that this doesn't mean that Pixelux's software will be free or open-source as Bullet is, but that it will allow licences to use the free and open-source Bullet library for a base level of physics and then to licence and layer Pixelux's DMM engine on top of that, and all of it accelerated on any GPU with an OpenCL driver.
Related Articles on Techworld
This is a clear shot across Nvidia's bow. Nvidia owns and licences the PhysX physics engine, having bought AGEIA early last year. PhysX supports all major consoles and PCs, but GPU acceleration on PCs is only provided through Nvidia's proprietary CUDA technology. AMD's goal is clearly to promote physics that can run on anyone's GPU or CPU, hoping that the combination of Bullet and Pixelux will produce top-notch physics solutions for developers that is equal or better in features and performance while making good business sense for developers.
Currently, Bullet is a distant third place among physics engines, behind the very popular Havok and PhysX products. What do you think? Will Bullet and Pixelux together with AMD put a dent in these other solutions? Will this move force Nvidia to port PhysX to OpenCL or DirectCompute?