Nobody gives a damn about mGPU unfortunately.
Faster interconnects and HSA type technologies will be needed to route around lazy developers. I was reading about a machine learning API from google earlier that allows you to pump code into it from a high level without having to explicitly target individual processors. So it seems more and more people understand the benefits.
Faster interconnects and HSA type technologies will be needed to route around lazy developers. I was reading about a machine learning API from google earlier that allows you to pump code into it from a high level without having to explicitly target individual processors. So it seems more and more people understand the benefits.