I would still highly recommend breaking it into solid and masked or translucent materials.
When the GPU renders opaque materials, it’s a simple matter of the GPU issuing a Write to the RAM and then continuing.
With Masked or translucent, the GPU has to also do a RAM Read of what the color underneath it is - it doesn’t matter if the source pixel is completely solid, it still does the RAM Read and Lerps with the new pixel with a 1.0 (testing it would just slow down the other pixels). A RAM Read has to issue the Read command, then wait for the data to be returned - so there is a Wait State.
There are intelligent caching routines built into the GPU, but the bottom line is that it will always be much slower than rendering an opaque material.
You may not notice with just a couple of objects on screen, but it will show up with more.
It’s good to cater to things like this right from the get-go.
Edit:
I just tested an RTX30 series card - and it’s nowhere near as bad as it used to be (rendering 31.6K rocks)
