The adaptive mode should be used when you have applied disproportional modifications to your mesh (such as by extensive use of Pinch brush). If your mesh deformation is fairly uniform, you’ll get faster results with adaptive mode off. Furthermore, you’ll be able to use the DPSubPix option to virtually subdivide the mesh to much higher resolution (even if your system can not normally handle such resolution).
Each DPSubPix level increases the virtual resolution of the mesh by four.
A value of 1 means it subdivides once, for 4 times as many polygons.
A value of 2 means it subdivides twice, for 16 times as many polygons.
A value of 3 means it subdivides three times, for 64 times as many polygons.
A value of 4 means it subdivides four times, for 256 times as many polygons.
Now let’s look at how many pixels your displacement map can have.
1024x1024 = about 1 million useable pixels
2048x2048 = about 4 million useable pixels
4096x4096 = about 16 million useable pixels
Finally, how many points does your model have at the highest subdivision level? What you want is to get a ratio of about 1 point for every texture pixel. So with a 4096x4096 map, you want to use a DPSubPix value that subdivides your model to 16 million points or less.
For a 1 million polygon model, that would be a DPSubPix of 2. If you use a value of 4 with a 1 million poly model, you would be forcing ZBrush to calculate 256 million polygons’ worth of detail, when the texture map can only hold 16 million pixels of data. That’s a serious waste of processing power and time.
