RAM chips are organized as x4 bits or x8 bits. On a memory module you're just
lining all these chips up in parallel so you have a 64bit wide data bus. All
512MB modules are made using 16 pieces of either 64Mbit x4, or 32Mbit x8. If
you use sixteen x8 chips you would get a 128bit data bus, so it has to be split
into two separate banks. Each with 64bit bus. That's why 512MB modules
based on x8 chips are dual bank. To build 256MB modules you can either use
sixteen 32Mx4 chips, or eight 32Mx8 chips. But you can't build a 256MB
module using 64Mx4 chips. To build 1GB modules you have to use thirty two
64Mx4 chips. Two 64Mx4 chips are stacked on top of each other with a lead
frame in between. Stacking basically converts two 64Mx4 chips into one
128Mx4 chip or one 64Mx8 chip. Then you stick 16 stacks onto a PCB.
Stacking also saves real estate, so you can fit more chips onto a shorter PCB
and keep trace lengths in spec.
If you are still unsure whether you are using x4 or x8 memory you can research the
memory chip by the SDRAM part numbers, or go to
search for jedec 21 c 4.20.4.