Uhh so turns out I wasnt using release compiler settings... In release mode with fastmath and whatnot, 4k renders in 5918ms with SAH bvh construction

4k render took 139887ms with 16 threads. 70584 primitives. If I use my improvement from the 240p image, it would take about 9 hours without bvh

240p Pre-bvh renders in about 5.5 minutes. With bvh 70584 primitives it renders in 1416ms