Connection Machines

The Connection Machine was a series of supercomputers that grew out of Danny Hillis's research in the early 1980s at MIT on alternatives to the von Neumann architecture of computation. He described the Connection Machine in AI Memo 646 (2MB PDF file) dated September 1981. The Connection Machines were programmed in a data parallel version of Lisp called *Lisp (prononced Star Lisp). The Allegro CL Examples and Utilities page has a Simulator for the *Lisp language.

CM-1

The CM-1, developed at MIT, was a "massively parallel" hypercube arrangement of up to 65,536 1-bit processors, each with 4K of their own RAM. A bit-serial data path was used to connect the processors. It's described in the book The Connection Machine, also by Danny Hillis.

CM-2

Shown with DataVault storage and visualisation monitor

Thinking Machines announced the CM-2 in April 1987. The memory per processor was increased and floating point units (FPUs) were added. It was reported to be capable of 2500 MIPS/2500 FLOPS (floating-point instructions per second). A 10 Gigabyte storage system called the DataVault was available.

Image © Thinking Machines Corporation, 1987. Photo: Steve Grohe.

CM-2a

The CM-2a is a low end machine in the CM-2 line. It has 4096 or 8192 processor elements.

Photo: Michael Ross

CM-5

Thinking Machines announced the CM-5 in November 1991. The CM-5 included a completely redesigned internal architecture communications, three separate networks for data, control, and diagnostics. It was capable of "performance in the range of 1 teraflops."

Photo: University of Maryland at College Park

Connection Machine links

Back to my symbolics page.
Back to my home page.