Abstract: We present two implementations of the Cube-4 volume rendering architecture on the Teramac custom computing machine. Cube-4 uses a slice-parallel ray-casting algorithm that allows for a parallel and pipelined implementation of ray-casting with tri-linear interpolation and surface normal estimation from interpolated samples. Shading, classification and compositing are part of rendering pipeline. With the partitioning schemes introduced in this paper, Cube-4 is capable of rendering large datasets with a limited number of pipelines. The Teramac hardware simulator at the Hewlett-Packard research laboratories, Palo Alto, CA, on which Cube-4 was implemented, belongs to the new class of custom computing machines. Teramac combines the speed of special-purpose hardware with the flexibility of general-purpose computers. With Teramac as a development tool we were able to implement in just five weeks working Cube-4 prototypes, capable of rendering for example datasets of 128³ voxels in 0.65 seconds at 0,96 MHz processing frequency. The performance results from these implementations indicate real-time performance for high-resolution data-sets.