Test scene 1

The source


#declare contrast=1;

camera{
location <0,1,0>
look_at 0}

plane{y,0.01

  pigment{
    gradient x
    pigment_map{
    	[0 rgbt 1]
	[0.5 rgbt 1]
	[0.8 rgbt 0]
	[1 rgbt 0]
	}
    scale 2
    translate <-1,0,0>
 
  }
}

plane{y,0 
   pigment{onion
	   pigment_map{
		   [0 rgb 1]
			 [0.5 rgb 1]
			 [0.5 rgb 0] 
			 [1 rgb 0]
			 }
		scale .02//*3
		}
	 finish{ambient 1}
	 
	 rotate y*34
	 }


-> Viewing Page (needs JavaScript)


Without Anti-Aliasing

Perfect


Test I - limited at about 4 samples per pixel

double size and scaling - 4.00 SPP

Method 1 (old) [+AM1 +A0.3 +R2 +J0] - 4.17 SPP

Method 2 (old) [+AM2 +A0.05 +R1 +J0] - 3.05 SPP


Recursion depth >1 resultated in SPP >6 so I had to keep it at 1.

Method 3 (like 2 but mostly using less SPP) [+AM3 +A1.5 +R2 +J0] - 4.75 SPP


Sample rates close to 4 could only be reached with a big threshold or recursion depth 1 (which would result exactly the same as method 2).

Method 4 [+AM4 +A0.5 +R2 +J0] - 4.51 SPP


Method 5 (like 4 but forcing midpoint always traced) [+AM5 +A1.2 +R2 +J0] - 4.40 SPP



Test II - limited at about 9 samples per pixel

triple size and scaling - 9.00 SPP

Method 1 (old) [+AM1 +A0.05 +R3 +J0] - 8.06 SPP

Method 2 (old) [+AM2 +A0.05 +R2 +J0] - 7.13 SPP


Again, recursion depth 3 resulted in SPP >14 so I had to keep it at 2.

Method 3 (like 2 but mostly using less SPP) [+AM3 +A0.8 +R3 +J0] - 9.11 SPP


Method 4 [+AM4 +A0.35 +R3 +J0] - 9.05 SPP


Method 5 (like 4 but forcing midpoint always traced) [+AM5 +A.45 +R3 +J0] - 9.09 SPP



Test III - wanting it perfect - limited at about 16 samples per pixel

4-times size and scaling - 16.00 SPP

Method 1 (old) [+AM1 +A0.05 +R4 +J0] - 15.43 SPP

Method 2 (old) [+AM2 +A0.05 +R3 +J0] - 15.29 SPP


Method 3 (like 2 but mostly using less SPP) [+AM3 +A0.05 +R3 +J0] - 15.09 SPP


Method 4 [+AM4 +A0.3 +R4 +J0] - 16.24 SPP


Method 5 (like 4 but forcing midpoint always traced) [+AM5 +A.05 +R3 +J0] - 10.8 SPP