Linking up
AMD has teamed up with its old pal, Cyberlink, to announce a broadening of the relationship between the pair and a strategic effort to focus on DirectCompute programming.
What with Cuda and Open CL stealing all the headlines lately, it's easy to forget that there is indeed a third way when it comes to speeding things up on a GPU, with Microsoft's DirectCompute programming interface and DirectX 11 toolkit gaining significant momentum in the run up to Windows 7 launch.
The announcement from AMD and Cyberlink wasn't exactly earth shattering in its implications, but the two firms did promise to work on accelerating Cyberlink's more compute intensive tasks like video transcoding, automated facial recognition and tagging, video editing and processing. AMD says it will also leverage its ATI Stream technology as well as the full specification capabilities of Win 7's DirectCompute API.
Of course, AMD didn't miss the opportunity to take a jibe at rival NVIDIA, bragging that its Radeon HD 5800 series was the only one to be fully DirectX 11 compliant. Not that one would especially need a high end Radeon to use the DirectCompute-enabled features on CyberLink, any GPU which is DirectX 10/10.1 works just fine.
Still, AMD reckons that running DirectCompute code on DX11 GPUs makes life easier for developers and allows for new algorithms for order independent transparency, ray tracing, better shadows, and depths of field.
