I just read this: http://www.nvidia.com/object/tesla_build_your_own.html
So after all, the "Nvidia Supercomputer" is actually nothing but a normal quad-core, with perhaps the GPU as an PCI add-on card. And all your referenced URL above are the Linux drivers to access the GPU card - ie, the OS is not running on the GPU. Thanks for the sharing!!! This is enlightening!!!
