With three LED Boards from Sureelectronics, a Arduino and my good old ikea couch table I build something new.
The displays are connected to the Arduino to share the row select pins (A,B,C and D) and the control lines EN, CLK & LATCH. The serial lines R1,R2,G1,G2 have their dedicated pins. So there are 19 pins needed out of possible 20 (not counting the reset pins as I/O). As the displays are loaded in 16 rows a 64bit some transformation has to be done while loading the display. This limits the update rate hard. A possible way to improve the frame rate is moving the transformation to the function setting the pixels.
Or maybe changing the connection scheme to common R1,R2,G1,G2 with separate EN & LATCH for each display which would result in 15 I/O pins needed but maybe suffer from slower update rate. As the display modules are chainable it is possible to connect them in series and load the 16 rows with 192bits each and using only 11 pins, but reducing the frame rate again. Since the ATMega328 cannot achieve high enough frame rates at any connection scheme I would suggest using the chained way.