Though the effects are regularly negligible, they can add up in given use cases, such as several water falls with thousands of droplets.
Is all the void space in the below image going to waste? Or is it culled from render? Does it come down to raw resolution of the images; a 16x16 a fourth the usage of a 32x32? Or is it more like image compression, where in the above image much of the center is compressed due to its similarity?
My intuition would tell me to use the bottom option, because although drawing the images is the negligible extent (except many up close overlapping >90%, yikes!), I summarize the rotation, and speed and such of an individual image is not.