try 12/0 vs 10/2. You want smaller tiles, even though there might be more, each tile is smaller (granularity). Being smaller means you can zero-in on exactly what tiles you want to render with less overhead.
Consider a larger tile where you can only see 1 pixel. You will have to check that entire tile vs the area/overhead of a smaller one; less overhead with more granular data = more performant.
Also, if your RVT covers your entire landscape, you likely can also uncheck clear-before-render since it’s always/only going to render that one thing, one time.