I have been working with this approach for so long i have become blind to all others. Originally it was too fast for the cost, thus the name. But then I made it so cheap the g to c/6 ratio is right around the recomended golden ratio. There is an uncomfortable amount of empty space in the instruction tape, so going faster is probably the way to do better.

The input atom is purified 5 times. The unused quicksilver goes into projecting a lead, copying the input metal. The first failed purification triggers an atom of quicksilver to be compared to the atom in arm 5. The timings of these quicksilvers encode the type of the input metal, and the previous metal.
If they arrive simultaneously, the counting metal is iterated.
If they arrive at different times, the counting metal and the copy of the previous input are added to the output.

The steady state rate is currently limited by arm 3.