GPGPU Gets Ready for Prime Time
Enthusiasts have been talking about general-purpose processing on graphics processing units (GPGPU) for about eight years now and some progress in that direction has been made in the last two or three years. Looks like in 2010 there will finally be critical mass of consumer applications that take advantage of GPGPU technologies.
Screenshot from vReveal by MotionDSP, a software that can take advantage of computing capabilities of Nvidia GeForce hardware
Until recently all the GPGPU-accelerated software was developed for specific hardware using specific tools, e.g. ATI Stream or Nvidia CUDA, there were no unification or standard-based approaches. As a result, the vast majority of GPGPU programs is exclusive to Nvidia CUDA architecture and work only on Nvidia GeForce graphics processors. However, since Microsoft DirectCompute and Khronos OpenCL standards were finalized in 2009 and appropriate software development kits are available from both ATI/AMD and Nvidia, standard GPGPU programs should emerge in 2010.
Thanks to industrial standards, we expect a lot of programs to finally start using graphics processing units for general-purpose computing, which will, without doubts greatly increase the importance of GPUs and will transform them into universal accelerators. This will help ATI, graphics business unit of Advanced Micro Devices, and Nvidia Corp. to boost sales of their advanced GPUs this year.