Questions about Memory Bus Width

[GangstaRas]

Okay, something has been bothering me on memory bus width and VRAM capacity. From my reading, it would appear that if a given GPU architecture is designed to use X amount of VRAM for a given bus, then going beyond that to 2X VRAM, it would cause a sharing of bus width lanes per memory chip. Someone was saying however that this VRAM would not be able to be used which completely confused me because if that's the case (assuming the fundamentals of RAM are the same) isn't that what we do everyday when we upgrade RAM? Put in more RAM on the same bus width? I need some clearing up for that part. And if I'm right, how then is performance affected because of this sharing? Is it hindered or when the memory is needed, it just gets called over the bus and the limitation is just the bandwidth of the memory. I'm presuming that no matter what application, gaming or some GPGPU processing, not all the memory is needed at once, idk though. Please clarify this.

The other thing is depending on what answer I get for the first question, if I had a GPU of the same architecture and same memory type, if one was 128 bit but it had crazy high memory clock speeds while the other is 256 bit but has slow clock speeds to give them effectively the exact same memory bandwidth, would there be an advantage of one bus width over the other regardless or is it just bandwidth that matters.