Apple doesn't release sytems with discrete GPUs to appease the Gaming Market.
They find a sweet spot that provides the best power consumption, price in volume, and provides the deepest and broadest capabilities it can leverage for it's OS and that means GCD/OpenCL/OpenGL in which more and more games are leveraging but in fact far more scientific, graphic and multimedia applications now expect to be optimized.
Nvidia continues to design it's GPGPUs around an optimized pipeline for CUDA. OpenCL has won this world and they are finally accepting it with their dump of NVidia PTX code stack into the LLVM/Clang project a few weeks back.
I'm building against it daily. The PTX Stack has come a long way, without Nvidia's help before they drop this source stack in trunk. AMD has been using LLVM/Clang for several releases to optimize their designs.
LLVM/Clang 3.1 is stamped.
LLVM/Clang 3.2 will finish off the C11 and C98 compliance needs and more goodies added by Apple and other third parties that will not only with FreeBSD moving to LLVM/Clang for versino 11 but more and more Linux platforms giving the option of their entire distributions built with LLVM/Clang or GCC as options.
All this is made possible thanks to Apple seeing and filling needs in the world where Hardware is not optimally being leveraged and more or less with GPUs having wasted over a decade pandering to Gaming as if its a real measure of what a GPGPU can do.
Games are benefiting but lets face it, games aren't the end all be all standard in driving graphics capabilities like it once did. More and more parallel programming is impacting everyone's lives, whether they know it or not. In 5 years people won't be waxing on about fps and 1080p as the bench standards, but photorealism without tearing in games and without taxing a system. Developers will have to do less trickery and consumers will only notice how pretty the textures are so long as their systems have discrete gpus on them.
Intel will have to copy AMD's APU approach sooner rather than later.
They find a sweet spot that provides the best power consumption, price in volume, and provides the deepest and broadest capabilities it can leverage for it's OS and that means GCD/OpenCL/OpenGL in which more and more games are leveraging but in fact far more scientific, graphic and multimedia applications now expect to be optimized.
Nvidia continues to design it's GPGPUs around an optimized pipeline for CUDA. OpenCL has won this world and they are finally accepting it with their dump of NVidia PTX code stack into the LLVM/Clang project a few weeks back.
I'm building against it daily. The PTX Stack has come a long way, without Nvidia's help before they drop this source stack in trunk. AMD has been using LLVM/Clang for several releases to optimize their designs.
LLVM/Clang 3.1 is stamped.
LLVM/Clang 3.2 will finish off the C11 and C98 compliance needs and more goodies added by Apple and other third parties that will not only with FreeBSD moving to LLVM/Clang for versino 11 but more and more Linux platforms giving the option of their entire distributions built with LLVM/Clang or GCC as options.
All this is made possible thanks to Apple seeing and filling needs in the world where Hardware is not optimally being leveraged and more or less with GPUs having wasted over a decade pandering to Gaming as if its a real measure of what a GPGPU can do.
Games are benefiting but lets face it, games aren't the end all be all standard in driving graphics capabilities like it once did. More and more parallel programming is impacting everyone's lives, whether they know it or not. In 5 years people won't be waxing on about fps and 1080p as the bench standards, but photorealism without tearing in games and without taxing a system. Developers will have to do less trickery and consumers will only notice how pretty the textures are so long as their systems have discrete gpus on them.
Intel will have to copy AMD's APU approach sooner rather than later.