To my knowledge, cards wouldn’t necessarily be a massive limiter when it comes to compute performance, although they do have other implications.
Cards let lumen cache expensive lighting behavior over multiple frames, and they do this by packing all the cards into a big atlas texture stored on the GPU. As you move around a scene, cards pop in and out so that the area around you always has coverage, up to certain user-defined limits. For very complicated objects, they inherently need more cards to get complete coverage, so they can be more expensive in how much cache real estate they take up.
So theoretically, if your scene is filled with many different opjects that would all be vying for surface cache coverage, there may be an advantage to reducing how many cards each object needs, but with how simple your objects are I’d be somewhat surprised if it was a big perf problem.