The Tang Primer 20K is the cheap target, and that is the point of it: it lowers the cost of having working hardware in more than one pair of hands, so bring-up is not bottlenecked on a single expensive board.
The vendor-specific parts stay confined to IP wrappers — HDMI and RAM — rather than spreading into the core or the bus fabric.
Results
The GW2A part closes at 36 MHz and draws 280 mW, using 17,440 LUTs and 5,113 registers. The demo is implemented and running on hardware.
It is the slowest and the hungriest of the three targets, which is the trade being made: it is also the one you can put on a desk for the price of a textbook.