min 4, max 32, using ray diffs and Cems filtering.

min 4, max 32. Hatlon, using ray diffs and Cems filtering, 11 seconds

Did my beer render with the new sampling method. It struggles with the normal map, probably because the variance becomes super high with minor changes in the normal map sampling. However, it maintains detail further in the background compared to before!

Same as before but with my mipmap implementation instead of the default filtering

Visualized Sample Counts

Same settings, but with default texture filtering using ray diffs

Halton, minimum samples was 2, max was 16, 720p 6 second render time. No texture filtering. 98 percent confidence for t values, with a deltamax of 0.01.