Academy Computer Science Systems architecture

Paper 1 · J277/01 · unit 1.1

Systems architecture

CPU purpose, Von Neumann, ALU/CU/cache/registers, fetch–execute, performance factors, and embedded systems. Exercises and a self-check quiz — nothing locked.

Everything on this course is open. Skip, jump, retry. Nothing is locked and there is no required order.

This unit is how a processor actually runs a program. You can read 1.1.3 first if you want — nothing here depends on finishing a quiz.

1.1.1 Architecture of the CPU

What the CPU is for

The CPU (Central Processing Unit) is the part of the computer that fetches instructions from memory, decodes them, and executes them. Everything else — screen, disk, network — is waiting on that loop.

OCR wants this purpose in one clean sentence: the CPU processes data and instructions by repeatedly fetching, decoding and executing them.

Von Neumann architecture

Almost every machine you use is Von Neumann:

  • One shared memory holds both instructions and data
  • A single bus moves them between memory and the CPU
  • The CPU is built from an ALU, a CU, registers, and cache

The bottleneck is that shared bus: you cannot fetch the next instruction and a piece of data in the same moment. That is the Von Neumann bottleneck. You do not need a dissertation on it — just the name and the cause.

The parts

Part Job
ALU (Arithmetic Logic Unit) Maths and logic: add, subtract, compare, AND/OR/NOT
CU (Control Unit) Decodes the instruction and coordinates the other parts — it sends control signals
Cache Small, very fast memory on or next to the CPU, holding copies of recently used instructions/data so the CPU is not always waiting on RAM
Registers Tiny, fastest storage inside the CPU, used during the current instruction

Cache is not “a register”. Registers are named slots with a job. Cache is a small pool of copies.

Registers you must know

Register Name Holds
PC Program Counter Address of the next instruction
ACC Accumulator Results of ALU operations; a working value
MAR Memory Address Register The address in memory we are about to read or write
MDR Memory Data Register The data or instruction just read from memory, or about to be written
CIR Current Instruction Register The instruction currently being decoded and executed

Exam trap: PC holds an address, not the instruction itself. The instruction lands in the MDR, then the CIR.

Fetch–execute cycle

One instruction, slowly:

  1. Address of next instruction is copied from PC → MAR
  2. That address is sent on the address bus; memory returns the instruction
  3. The instruction is copied into the MDR, then into the CIR
  4. PC is incremented so it now points at the next instruction
  5. CU decodes the instruction in the CIR
  6. Execute: ALU might calculate; memory might be read/written via MAR/MDR; ACC might update
  7. Repeat

Some questions say fetch–decode–execute. Decode is the CU’s step. Same cycle.

If the instruction is a jump, execute overwrites the PC with a new address instead of “just going to the next one”.

Buses (enough for this unit)

  • Address bus — CPU tells memory where (typically unidirectional)
  • Data bus — the actual bits, both ways
  • Control bus — read/write and timing signals

You will meet these again in networks and in “what limits performance”.

1.1.2 CPU performance

Three factors OCR hammers. Learn the direction of the effect and a limitation.

Clock speed

The clock ticks millions or billions of times a second (Hz, MHz, GHz). One tick is a slice of time the CPU can use to move through fetch–execute.

  • Higher clock speed → more cycles per second → more instructions in the same time (if nothing else is waiting)
  • Hitting the ceiling: heat, power, stability. You cannot infinitely overclock a laptop in an exam answer.

Cores

A core is a processing unit that can run its own fetch–execute cycle.

  • More cores → more instructions in parallel if the software can split the work
  • A single-threaded program may barely notice a second core
  • Dual-core is not “twice as fast at everything”

Cache size

Larger cache → more of the working set sits next to the CPU → fewer slow trips to RAM → higher effective performance.

Cache is still small and expensive. “Just make cache as big as RAM” is not an exam answer.

How to write a compare question

“Explain how increasing clock speed and increasing cache size can improve CPU performance.”

Give each factor its own sentence: what changes, why that helps fetch–execute, one limitation. Do not mash them into one blob.

1.1.3 Embedded systems

An embedded system is a computer designed for one (or a small set of) dedicated functions, usually inside another device.

Examples worth having in your pocket: washing machine controller, car engine management, digital watch, dishwasher, a smart thermostat, a traffic light controller.

Typical characteristics (pick the ones that fit the question):

  • Dedicated task, not a general-purpose laptop
  • Often real-time (must respond by a deadline — airbag, ABS)
  • Limited OS or none you would recognise
  • Built for low cost, low power, reliability
  • Program often in ROM; user cannot install games on the toaster

Not embedded just because it is small: a smartphone is a general-purpose computer that happens to fit in a pocket.

When would you not use one? When the device must run arbitrary software, be upgraded like a PC, or do many unrelated jobs.


Jump to the exercises and quiz below whenever you like — or open another unit from the sidebar.

Practice · optional · answers on this page

Exercises

Do as many or as few as you like, in any order. Hints and a model answer sit under each task.

1.1.1 · e1 paper

Name the register

For each sentence, write PC, ACC, MAR, MDR or CIR. 1. Holds the address of the next instruction to be fetched. 2. Holds the instruction currently being decoded. 3. Holds the address of the memory location being read or written. 4. Holds data copied from memory, or data about to be written to memory. 5. Holds the result of an addition just done by the ALU.
Hint
Addresses live in PC and MAR. The instruction itself ends up in CIR. ACC is a working value.
Show a model answer
1. PC 2. CIR 3. MAR 4. MDR 5. ACC
1.1.1 · e2 paper

Fetch–execute in order

Write these steps in the order they happen for a normal (non-jump) instruction: A. PC is incremented B. Instruction copied into CIR C. Address copied from PC to MAR D. CU decodes the instruction E. Memory contents copied into MDR F. Instruction is executed
Hint
You cannot decode what you have not fetched. PC usually ticks after the instruction is safely in the CPU.
Show a model answer
C → E → B → A → D → F (Some mark schemes swap the exact moment PC increments with copying into CIR — both “after fetch, before execute” is the idea. If you put A immediately after C, say why: some textbooks increment as soon as the address has been copied.)
1.1.2 · e3 paper

Three factors, three sentences

A games console is upgraded from 2 cores at 1.5 GHz with 1 MB cache to 4 cores at 2.5 GHz with 4 MB cache. Write three short paragraphs (clock, cores, cache). Each must say what changed, how that can improve performance, and one reason it might not double the frame rate.
Hint
Parallelism needs software that can use cores. Clock hits heat. Cache still is not RAM.
Show a model answer
Clock: 1.5 → 2.5 GHz means more cycles per second so more instructions in the same time, but heat/power limits and RAM waits still cap real speed. Cores: 2 → 4 can run more fetch–execute cycles at once if the game is written to split work; a single-threaded part will not go twice as fast. Cache: 1 MB → 4 MB holds more recently used instructions/data on-chip, fewer RAM trips; it is still tiny compared with main memory so misses still happen.
1.1.3 · e4 think

Embedded or not?

Say embedded or general-purpose, and one reason: 1. Digital microwave controller 2. School laptop 3. ABS braking computer in a car 4. Smartphone 5. Traffic-light controller at a junction
Hint
Dedicated function inside another device vs “install whatever you want”.
Show a model answer
1. Embedded — dedicated cooking program, inside the microwave 2. General-purpose — runs many applications the user chooses 3. Embedded — real-time dedicated safety function in the car 4. General-purpose — full OS, arbitrary apps, even though it is small 5. Embedded — dedicated control of lights, not a desktop
1.1.1 · e5 paper

Exam-style 4 marks

Describe the purpose of the ALU and the CU in a CPU. (4)
Hint
Two marks each if you give purpose + a concrete example of what it does.
Show a model answer
ALU: performs arithmetic (e.g. addition) and logical operations (e.g. AND/OR/compare) on data in registers such as the accumulator. CU: decodes instructions in the CIR and sends control signals that coordinate the ALU, registers and memory so the instruction is actually carried out.

Check yourself · not a gateway

Quiz

Mark it, reveal it, or skip it. A low score does not close anything. Try again as often as you want.

1 1.1.1 Which register holds the address of the next instruction to be fetched?
2 1.1.1 In Von Neumann architecture, what is shared between instructions and data?
3 1.1.1 Which of these are registers in the OCR J277 CPU model? (Select all that apply)
4 1.1.1 During fetch, the address of the instruction is first copied from the PC into the…
5 1.1.2 Increasing the number of cores improves performance mainly because…
6 1.1.2 Why does a larger cache often improve CPU performance?
7 1.1.3 Which device is the best example of an embedded system?
8 1.1.3 A smartphone is an embedded system because it is small and battery-powered.