GPU based vs Ray Casting Only Volume Rendering


Currently I’m working with a volume data using paraview 5.6.0 release +cluster remote rendering (multiple nodes + multiple gpus). Before hand I have to say the data set is a tiff stack of 4k textures of size 60k x 60k x 600.

I am confused on the different results I’m getting when switching the volume rendering mode between GPU based and Ray Casting Only.
This first image is the data rendered using the former. The scene looks patched (I assume is due to the parallel rendering), some parts have more colors than others. Additionally, I am getting more pixels of interests related to the color/opacy transfer function.

This second image is using the latest. The image is not patched, but I have less pixels of interests.

The funny thing is, using ‘Ray Casting only’ looks similar to what we are expecting to see form the data set, and the ‘GPU based’ looks like there’s too much noise around the area of interest.

So, I dont know which rendering mode is the right way to continue using for further investigations. Also, What’s the main difference on using smart mode? Furthermore, using OSPRay makes all the data set completely transparent. Any suggestions?

Thanks for the help.