In previous generations when they've had differing memory sizes for the same card, that was the ONLY change. So, it was useful for something like CUDA, but usually not for gaming. A specific audience.
For the 4080, the 12GB version has the following changes:
* 12GB VRAM vs. 16GB VRAM (the obvious one from the name)
* 7,680 CUDA cores vs. 9,728 CUDA cores for the 16 GB.
* 192-bit memory bus vs. 256-bit memory bus (understandable, since this scales with memory size... but also probably means the memory itself is slower).
This isn't just a different amount of memory, it is fundamentally a different product and should be marketed as such. Instead it's Nvidia being greedy.