And this is why I worry about the new MacPro. OpenCL acceleration is not nearly as fast as CUDA on the PC, and the Mac has always had a far inferior OpenCL installation. Multi GPU’s are incredibly hard to code for and only give moderate performance enhancements on even the most mutli-GPU aware games on the PC, so they are only for the most hard core of gamers. The only exception to this is a Maximus configuration from NVIDIA which is a Quadro with a TESTLA card (the non-consumer version of a Titan), which has incredible power and speed and really can use both processors. This all leaves the new MacPro in the code. Most software won’t be coded for multiple AMD GPU’s and even if they are the performance increases are usually pretty modest, and OpenCL can’t touch CUDA! So why made a new “pro” machine without the option for CUDA? The only argument I can see if form over function, and that seems to be what the new MacPro is all about!