The high throughput example can saturate the backplane connection on a 7976 if you change a few things on the block diagram. Specifically the number of samples that are put into the DMA fifo every tick as well as the constant that reports that maxmimum throughput.
If you don't want to modify that you can take a look at some demo code I whipped up a while back. Set the 'duty rate' equal to the 'clock rate' to get the full 3.2GB/s.
