1) Task Modeled as Scalar Repetition, Not True Throughput
If the task is broken into many small, equal steps (say, simulate a million identical operations), a processor with a smaller time per operation will win in that model, even if real‐world performance would favor a more advanced CPU with pipelining, wider buses, and more instructions per cycle.
2) No Memory Bandwidth or Instruction Set Overhead Included
The 68000 has 32‐bit internal registers and a rich instruction set, which adds overhead in simple repetitive tasks. The Z80, being an 8‐bit CISC processor, often performs simpler tasks faster in extremely constrained or unrealistic scenarios (e.g., single‐threaded loops with limited memory).
The Motorola 68000 is objectively far more powerful:
32‐bit internal architecture
More addressable memory (24‐bit address bus = 16 MB)
Richer instruction set and support for more complex operations
The Zilog Z80:
Is an 8‐bit CPU with a 16-bit address space (64 KB max)
Has fewer instructions per cycle and limited addressing modes
But does well in small embedded tasks and legacy systems
So the Z80 outperformed the 68000 in your test because:
The test valued simplicity and fast loop iteration over instruction complexity.
Your spreadsheet likely gave it a better "time per calc" value.
The task wasn't optimized to use the 68000's strengths (e.g., handling large data, multitasking, efficient memory access).