← 🚰 Instruction Pipelining & Hazards
Subtopic 2 of 8

2.2Cycle time

The idea

All stages share one clock, so the slowest stage decides the speed of the whole pipeline.

Think of it like this: An assembly line moves at the pace of its slowest worker; everyone else waits.

Why it exists

It explains why real speedup is below the number of stages and why designers try to balance stages.

How it works

  1. Clock period = longest stage delay + latch (register) overhead.
  2. Unpipelined time per instruction = sum of all stage delays (no latch).
  3. Speedup (many instructions) ≈ sum of delays ÷ clock period.
  4. Shortening a non-bottleneck stage changes nothing.

Worked example: Delays 200, 250, 150, 300, 200 ps + 20 ps latch

Clock = 300 + 20 = 320 ps. Unpipelined = 1100 ps. Speedup ≈ 1100 ÷ 320 = 3.44, not 5.

Now play with it

  • →Drag the MEM (slowest) slider down and watch the clock fall. Drag a fast stage and watch nothing improve.
IF
ID
EX
MEM
WB

Green = stage work (ps), red = idle waiting for the slowest stage.

Pipelined clock (max + latch)320 ps
Non-pipelined time/instr1100 ps
Max speedup (n → ∞)3.438
Ideal speedup (equal stages)5

Check yourself

Stage delays are 4, 6, 5 and 8 ns and the latch adds 1 ns. What is the pipelined clock period?

Remember: Clock = slowest stage + latch.
Exam notes (the dense version)

Clock = slowest stage + latch (register) delay. Unbalanced stages waste time.

Previous year questions: Instruction Pipelining & Hazards (15)

PYQs are grouped by topic, so this is the full set for Instruction Pipelining & Hazards; try the ones that match this subtopic. Open all 15 on the topic page.