Graphics Product Group (Part 3)

Stream Computing is AMD's general purpose graphics processing initiative, where the arithmetic prowess of GPUs is harnessed to accelerate applications, in fields of scientific research, gaming, productivity, and entertainment. While AMD doesn't have a proprietary software backbone for its GPGPU technology, such as CUDA for NVIDIA, it relies on OpenCL to run its GPGPU computing layer. Applications will then take advantage of the parallelism stream processors provide, and offload the CPU. The December release of the ATI Catalyst Software suite has been announced to contain GPGPU software features. AMD will be releasing the ATI Avivo HD Transcoding software for free to all users of compatible Radeon graphics accelerators. Additionally, there are a lot of software releases lined up that harness the capabilities of Radeon GPUs. It could also be noticed that AMD is pushing forth the "Balanced Computing" initiative, where Phenom processors handle serial loads of computing, while Radeon GPU handle parallel loads. Reading between the lines, it could be seen that the GPGPU is projected to make up for performance shortfalls of the Phenom processors, with some of the most CPU-intensive tasks such as transcoding. AMD slipped in an interesting slide, where it expects its platform to be twice as fast with transcoding, as that of NVIDIA, and over 12x faster than a single Intel processor-only platform, at a third of the cost. Sounds ambitious.