Amdahl's Law is the principle that a program's speedup from parallel computing is limited by its sequential portion, the part that must run one operation at a time. No matter how many processors you add, total time can never drop below the sequential portion, so extra parallelism gives diminishing returns.
Amdahl's Law answers a simple question. If I throw more processors at a program, how much faster can it actually get? The answer depends on how much of the program must run sequentially. Every parallel solution has two pieces (EK CSN-2.B.1): a parallel portion you can split across processors, and a sequential portion that runs one step at a time no matter what. You can shrink the parallel portion's time by adding processors, but the sequential portion never shrinks. It sets a hard floor on total runtime.
Here's the intuition. Think of a group project where one teammate has to write the intro alone before anyone else can start. Adding 50 more teammates speeds up the group sections, but the project can never finish faster than that solo intro takes. That's why the CED says that as you increase parallel computing, the benefit of adding each additional processor gets smaller and smaller (EK CSN-2.B.5). The maximum theoretical speedup equals the total single-processor time divided by the sequential time. If a task takes 25 seconds total and 5 of those seconds are sequential, the best possible speedup is 25 ÷ 5 = 5x, even with infinite processors.
Amdahl's Law lives in Topic 4.3 (Parallel and Distributed Computing) in Unit 4: Computer Systems and Networks. It directly supports learning objective 4.3.A, which asks you to compare solutions and determine their efficiency, and 4.3.B, which asks you to describe the benefits and challenges of parallel computing. The CED never names "Amdahl's Law" outright, but EK CSN-2.B.5 is Amdahl's Law in plain English. It says when you increase parallel computing, the benefit eventually levels off because of the sequential overhead. So when the exam asks why doubling your processors didn't double your speed, Amdahl's Law is the answer it's looking for, even if the question never uses the name.
Keep studying AP® Computer Science Principles Unit 4
Sequential Computing (Unit 4)
Sequential computing, where operations run in order one at a time (EK CSN-2.A.1), is the villain in Amdahl's Law. The sequential portion of a program is the part that refuses to be parallelized, and its runtime is the floor your total time can never go below.
Speedup and Efficiency Comparisons (Unit 4)
The CED measures efficiency by comparing how long solutions take on the same task (EK CSN-2.A.4). Speedup is sequential time divided by parallel time, and Amdahl's Law tells you the ceiling on that number before you even run the program.
Distributed Computing (Unit 4)
Distributed computing spreads a program across multiple devices (EK CSN-2.A.3) to solve problems too big for one computer. Amdahl's logic still applies, and distributed systems add a second tax in the form of communication time between devices, which acts like more sequential overhead.
Scalability of Parallel Solutions (Unit 4)
EK CSN-2.B.2 says parallel solutions scale better than sequential ones, and Amdahl's Law is the fine print on that claim. Scaling works great at first, then flattens out as the sequential portion dominates. Both ideas are true at once, and the exam likes testing whether you know that.
This shows up as multiple-choice questions on the AP CSP exam, usually in two flavors. The first is conceptual, asking which statement best describes why adding processors gives diminishing returns. The correct answer always points to the sequential portion as the limiting factor. The second flavor is computational. You get a sequential time and a parallel time, like a 5-second sequential portion and a 20-second parallelizable portion, and you compute the maximum theoretical speedup. Take the total single-processor time (25 seconds) and divide by the sequential time (5 seconds) to get 5x. Another common version states the percentage instead, like "20% of the code cannot be parallelized," which also caps speedup at 1 ÷ 0.20 = 5x. The trap answers assume infinite processors mean infinite speedup. They don't, and that's the whole point of the law.
Speedup measures how much faster a parallel solution actually ran. You divide the sequential solution's time by the parallel solution's time. Amdahl's Law is about the maximum possible speedup, the theoretical ceiling set by the sequential portion. On the exam, read carefully whether you're computing an actual speedup from two given times, or the best-case speedup assuming unlimited processors. They use similar division but answer different questions.
Amdahl's Law says the sequential portion of a program limits the total speedup you can get from parallel computing, no matter how many processors you add.
Maximum theoretical speedup equals total single-processor time divided by sequential time, so a task with 5 sequential seconds out of 25 total can speed up at most 5x.
Adding more processors gives diminishing returns because each new processor only helps the shrinking parallel portion, never the sequential portion (EK CSN-2.B.5).
A parallel program's runtime can never be shorter than its sequential portion, since that part must run one operation at a time.
The AP CSP exam tests this idea through MCQs in Topic 4.3, often by giving you sequential and parallel times or a percentage that cannot be parallelized.
Parallel solutions still scale better than sequential ones (EK CSN-2.B.2); Amdahl's Law just explains why that scaling eventually flattens out.
Amdahl's Law is the principle that a program's speedup from parallel computing is capped by its sequential portion, the part that must run one operation at a time. In the AP CSP CED, this idea appears as EK CSN-2.B.5 under Topic 4.3, which says adding more parallel computing yields less and less benefit.
Divide the total time on a single processor by the sequential portion's time. If a program takes 25 seconds total and 5 seconds are sequential, the maximum speedup is 25 ÷ 5 = 5x. If you're given a percentage instead, like 20% non-parallelizable, divide 1 by that fraction (1 ÷ 0.20 = 5x).
No, and this is the most common trap answer on the exam. Extra processors only speed up the parallel portion, so once that portion is tiny, more processors barely help. Total runtime can never drop below the time the sequential portion takes.
Speedup compares two actual runtimes (sequential time divided by parallel time) to see how much faster a solution ran. Amdahl's Law predicts the theoretical best-case speedup assuming unlimited processors. One describes what happened, the other describes the ceiling on what's possible.
Not by name, but the concept is fully tested. EK CSN-2.B.5 states that increasing parallel computing gives diminishing benefits because of the sequential overhead, which is Amdahl's Law in plain English. Exam questions test the idea whether or not they say the name.
Connect this key term to the AP exam workflow: review the course, practice questions, and check related study tools.
Review units, study guides, and course resources.
Check this vocabulary in multiple-choice context.
Apply key concepts in written AP responses.
Estimate the exam score you are working toward.
Review the highest-yield facts before practice.
Put the full course together before test day.