After the fact, it seemed - even to us - that there was only one natural, obvious form for the Connection Machine CM-1/CM-2: the cube-of-cubes that reflected the internal 12-dimensional hypercube architecture of the machine. Of course at the beginning of the design process, things were not clear at all.

The first problem Danny asked me to solve was whether it was physically possible to build the machine. After talking with the electronic hardware engineers, the answer was "yes, but it's pretty big." These were the first feasibility study drawings I did based on their information:

Hypercube wiring:

The processor boards would be 2'x2', making the machine ~ 10' tall x 5' wide x 4' deep:

That was unacceptably large, so the electrical engineers went back to their drawing boards and reduced the board size considerably. After that, the machine would have fit into a normal 19" electronic equipment rack. But that's not what we wanted.

This is the point, in February 1984, when Allen Hawthorne and Gordon Bruce came on board to work on the industrial design (see their interviews on the "Design Legacy" page). We described the cube-of-cubes structure of the machine to them, but I didn't want to lock us into a cube structure without looking at alternatives. With architect Tom Chytrowski I explored various forms, as did Al and Gordon. But each of us drew our versions of a cube-of-cubes as well, and when we came back together to view all the sketches, it was clear that this was the right form for the machine.

Detail photos from a real machine (pardon our dust!), including details that Allen and Gordon talk about in their video interviews (see Design Legacy page).