Yes, it's for A8 only. The A8 renderer is much faster than the A7 renderer, but requires the whole sprite image in video memory. A7 split the sprite in separate bitmaps and rendered them separately.

You can emulate the A7 behavior by script: split the image in separate BMAPs with bmap_createpart(), and animate the sprite with the bmap_for_entity() command. Then you have the old slow A7 behavior, but can step around your 3D card size limit.