The following is Figure 1 in Gene M. Amdahl's "Validity of the single processor approach to achieving large scale computing capabilities" (1967) (PDF):
The text describes this as representing the performance (presumably vertical) as a function of the fraction of instructions that can be executed in parallel for three machines. Machine A has 32 arithmetic execution units, machine B has 3-deep pipelined execution units working on 8-element vectors, machine C has 3-deep pipelined execution units with 8-wide scalar execution (if I understood the descriptions correctly). With a note that "The probable region of operation is centered around a point corresponding to 25% data management overhead and 10% of the problem operations forced to be sequential."
How did Amdahl derive this graph and how is the graph meant to be interpreted?