What's the relation between many-core and gpgpu? parallelism

Hi, microsoft lead a workshop on manycore:
http://science.officeisp.net/ManycoreCompu…tionPapers.aspx
they say manycore is facing big challenges from language/runtime/automation/programming model/…etc.
While CUDA team has done and is doing an excellent job, I couldn’t ask if there’re relationships between the manycore world and GPGPU world.
Any comments are welcome! Thanks.