Increasing the samples decreases the "graininess", at the cost of performance Much Faster Link: https://turbowarp.org/1143280388/fullscreen
Thanks to Dr. Peter Shirley for the formulas