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

6.4Amdahl's law

The idea

Amdahl's law: speeding up only part of a task is limited by the part you did not speed up.

Think of it like this: Making the highway ten times faster does not help if half your trip is a traffic-light town.

Why it exists

It tells you where optimisation effort is worth it.

How it works

  1. f = fraction of time that gets enhanced; s = speedup of that part.
  2. Speedup = 1 ÷ ((1 − f) + f ÷ s).
  3. Even with s = ∞ the speedup is only 1 ÷ (1 − f).
  4. So the unenhanced fraction rules.

Worked example: f = 0.7

Even with an infinitely fast enhancement, max speedup = 1 ÷ 0.3 = 3.33.

Now play with it

  • →Slide f up and watch the ceiling of the curve rise; press Sweep and see how it flattens.
1100speedup s of the enhanced part3.60overall speedup

s = 10.0 → overall 2.70x (limit 3.33x)

The curve flattens at 1/(1−f): the part you did not speed up dominates.

Check yourself

70% of execution time is sped up by an infinite factor. Maximum overall speedup?

Remember: The part you cannot improve sets the limit.
Exam notes (the dense version)

Speeding up fraction f by s overall gives diminishing returns; bound = 1/(1−f).

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.