← 📈 Performance: CPI, MIPS, Amdahl
Subtopic 2 of 4

6.2MIPS isn't everything

The idea

MIPS (millions of instructions per second) is an easy number but a poor yardstick, since different machines need different numbers of instructions for the same program.

Think of it like this: Comparing cars by steps per second rather than distance covered.

Why it exists

Only execution time on the same program tells which machine is faster.

How it works

  1. MIPS = clock ÷ (CPI × 10⁶).
  2. A machine with fewer, more powerful instructions may have a lower MIPS yet finish sooner.
  3. So always compare execution time.

Worked example: Machine A vs B

A: 1000M instrs, CPI 1: 2000 MIPS, 0.5 s. B: 400M instrs, CPI 2: 1000 MIPS, 0.4 s. B is faster with half the MIPS.

Now play with it

  • →Edit the instruction counts and watch "faster machine" and "higher MIPS" disagree.
A: MIPS2000
A: time (s)0.500
B: MIPS1000
B: time (s)0.400
Faster machineB
Higher MIPSA

With the defaults A has twice the MIPS but B finishes sooner: MIPS ignores how many instructions the program needs.

Check yourself

Does a higher MIPS rating always mean a faster machine for a given program?

Remember: Compare execution time, not MIPS.
Exam notes (the dense version)

MIPS can mislead: different ISAs need different IC. Compare machines by execution time.

Previous year questions: Performance: CPI, MIPS, Amdahl (3)

PYQs are grouped by topic, so this is the full set for Performance: CPI, MIPS, Amdahl; try the ones that match this subtopic. Open all 3 on the topic page.