Which type of collider is better: Tilemap or sprite?

I see people using a sprite object as a collider, but this generates dozens of instances to cover the whole scenario with sprite collisions (for example, a blue block).

Already if I use another collision-only tilemap with solid behavior, I will have two tilemaps in the scene, but the collision tilemap I put in a specific layer.

Which is better to make the collisions of the scenario? Which one performs the best?

In the construct 3 home screen demo game a sprite is used, with over 30 instances in the phase to cover the whole scenario. Is this how it is done? Or better to use a tilemap just for that?

In general, answer your own performance questions with measurements.

And as usual, you probably won't be able to measure a difference, because optimisation is a waste of time until you have measurements identifying a specific thing that needs optimising.

Thank you Very much.I actually used the debugger and consumed less performance with tilemap. And it's also faster to make a tilemap part for collisions on a separate layer. I don't know why ex. of platform game from the homepage was made with sprites, generating many instances. Maybe you have better control over the placements, I don't know. but thanks