NVIDIA has had two exclusive features to pull users to its graphics cards, PhysX for accelerated graphics effects and CUDA for GPGPU applications. During GPU Technology Conference NVIDIA announced that together with PGI it will developed an x86 version of CUDA.
By enabling CUDA code to run on x86 processors NVIDIA opens up to an enormous market for CUDA-based applications, but at the same time limits the overhand of its own solutions. The graphics card maker was clear that its own GPU solutions will still offer the best performance. NVIDIA will developed the new version together with PGI (The Portland Group), but we don’t know when it will be ready.
Now that Intel bakes in a video processor into Sandy Bridge there is every reason for NVIDIA to revise its alternatives for spreading the CUDA brand. Even if the great earnings are on the server/HPC market, where CUDA is a weapon to be recognized. The great question is how great the difference will be between CPU and GPU accelerated CUDA applications.
