The link is incredibly light on actual content but this seems to be good news for AI enthusiasts as perhaps now we'll get a reasonable competitor to CUDA/CUDNN and their associated hardware for running GPU accelerated machine learning. Intel seems to be taking the ML/AI space seriously and this move seems very likely to be related. Yes I'm aware of OpenCL as I am also aware of it's level of support with libraries such as PyTorch, Tensorflow, Theano -- it isn't the first class citizen that CUDA is. While those libraries aren't perfect they offer the experience of writing the experiment on your laptop without a GPU, validating, then running the full experiment on larger hardware.
In my ideal world competition from intel would force NVidia to play nice with OpenCL or something similar, and encourage competition in the hardware space instead of driver support space. Unfortunately the worst-case looks something more like CUDA, OpenCL and a third option from Intel with OpenCL like adoption. :(
In my ideal world competition from intel would force NVidia to play nice with OpenCL or something similar, and encourage competition in the hardware space instead of driver support space. Unfortunately the worst-case looks something more like CUDA, OpenCL and a third option from Intel with OpenCL like adoption. :(