Hey GabenJr,

Figured I'd drop you a line, you seem to be interested in this stuff.

 I'm at a research school in theoretical physics these two weeks, and spent one evening talking to a professor about scientific computing. He said that quite a few people are moving to Xeon Phi's rather than GPUs. The reason seems to be primarily that the time invested in programming is significantly less, since methods like OMP easily make use of the extra flops. With videocards, you need to make dedicated CUDA code. That seems to be the primary difference - and this investment in a 6k Xeon Phi is easily earned back by having developers spend less time coding CUDA. 


A GPU/XeonPhi/CPU parallelism video would be something a few of us are interested in (although, I do not know how many and if it is sufficient to make it interesting for LMG).

Just thought you'd like to know.