When someone mentioned 3d transistors at first i got this idea, that in the end turned out to not be what 3d transistors are like.
First lemme describe the "cell":
A cell, in this context, is a component, or group of components, with six inputs/outputs arranged like the 6 faces of a cube that behaves in the following manner:
Each pair of opposing faces are connected, but only let electricity thru based on the voltage differential of one of the other pairs (each pair obeys to one other pair and controls the remaining pair), allowing electricity in one direction if the voltage difference in it's respective commanding pair is positive, in the opposite direction if the difference is negative and in no direction if there is no voltage difference (within a threshold if the properties of the material or other components require it).
And now to define the "computing block":
A computing block is formed by an array of cells arranged in a cubic lattice, with all cells oriented the same way (each cell's X,Y and Z axis aligned and the respective +'s and -'s pointing the same direction), with arbitrary dimensions in each axis.
With some sort of external logic for routing the signals in and out (such that the routing can be changed at runtime), would a computing block such as described above actually be capable of producing meaningful results given the correct input (input being the routing patterns and trinary signals, negative current, positive current and zero current) to the point of being considered turing complete? How could the cells be built? Is there such a component already made or an already known combination of components that produces such a behavior? How would the routing logic work and how could it be built? Would the dimensions matter or algorithms for the inputs can be adapted to work on an a parallelepiped of any dimension bigger than a minimum? Could the output of the block be used for controlling the routing logic or additional elements would be needed to interpret the output and adjust the routing?