I.5 Components of a Computer System

⬆️ Back to: P-I.I. Introduction to Computing

🧭

Page Outline — click a heading to jump. Past-paper questions sit in the last section.

🎯

Exam pattern (2016–2026): 2017 Q2(b) Harvard vs Von Neumann (6) · 2026 Q2(c) hardware components (8).

---

1. CPU — the "Brain"

1.1 Control Unit (CU)

• Role: directs and coordinates everything.
• Runs the cycle: fetch instruction → decode it → send data to ALU or memory.
• Does not calculate. Think traffic controller, not calculator.

1.2 Arithmetic Logic Unit (ALU)

• Arithmetic: add, subtract, multiply, divide.
• Logic: AND, OR, NOT, XOR, comparisons (>, =).
• Only unit that produces results. Every spreadsheet formula and if ends up here.
• Multi-core CPUs: one ALU per core, so calculations run in parallel.

1.3 Registers

• What: tiny, fastest storage inside the CPU chip.
• Key registers:
  • Accumulator (ACC): intermediate results
  • Program Counter (PC): address of next instruction
  • Instruction Register (IR): instruction being decoded
  • MAR / MDR: memory address and memory data
• Limit: only a few, each holding one word (32 or 64 bits).

1.4 CU vs ALU

• 🧠 Formula: the CU directs, the ALU computes.

2. Primary Memory

2.1 RAM

• Volatile, read-write.
• Holds: running programs, active OS parts, their data.
• Random access: any location reached in about equal time.
• Lost on power cut — why unsaved work disappears.
• More RAM = more programs running smoothly.

2.2 ROM

• Non-volatile; holds firmware such as BIOS/UEFI.
• Variants:
  • PROM: programmed once
  • EPROM: erased by UV light
  • EEPROM / flash: erased electrically; used in modern firmware
• Guarantees the computer can always find its start-up instructions.

2.3 Cache

• Small, very fast memory between CPU and RAM.
• Works by locality of reference: programs reuse recent data.
• Levels:
  • L1: smallest, fastest, inside each core
  • L2: larger, a bit slower
  • L3: largest, shared by all cores

3. Secondary Storage

• Non-volatile, large, slower than RAM.
• CPU cannot use it directly; data is first loaded into RAM.
• HDD: magnetic platters; cheap and large, but slow and fragile.
• SSD: flash chips, no moving parts; fast and tough.
• Optical (CD/DVD/Blu-ray): laser-read; portable, limited capacity.
• USB flash / memory cards: small and portable; lower capacity.
• Full comparison: I.9 Storage Media.

4. Memory Hierarchy

• Order (fastest → slowest): Registers > Cache (L1 > L2 > L3) > RAM > SSD > HDD > optical/tape.
• Going down: speed falls, capacity rises, cost per byte falls.

• Scale: a 10 ms HDD access is about 10 million times slower than a register.

5. Input/Output Devices

• How data enters and results leave the system.
• Details: I.7 Input Devices and I.8 Output Devices.

---

6. ⚠️ MCQ Traps

• CU directs; ALU computes.
• Registers are faster than cache.
• CPU cannot process directly from secondary storage.

7. ✅ Self-Test Questions (with Answers)

Q1. What are the two sub-units of the CPU, and what does each do?

• Control Unit: fetches, decodes, sends control signals; does no calculation.
• ALU: arithmetic and logic operations; produces results.
• Registers: third element — ACC, PC, IR, MAR/MDR.
• Add the table in 1.4.

Q2. Rank registers, cache, RAM and secondary storage from fastest to slowest.

• Registers > Cache > RAM > Secondary storage.
• Reason: speed vs capacity vs cost trade-off.
  • Registers sit inside the CPU (one clock cycle).
  • Cache is SRAM; RAM is DRAM; storage is behind an I/O controller.
• Add the table in section 4.

Q3. Why can't the CPU process data directly from secondary storage?

• Addressing:
  • RAM is byte-addressable on the address bus.
  • Disks are block-addressable (512 B or 4 KB) behind SATA/NVMe controllers.
  • The CPU can only request a block transfer into RAM.
• Speed:
  • At 3 GHz, a 10 ms disk wait = about 30 million idle cycles.
• Role: storage keeps data permanently; RAM holds the working set.
• Actual flow:
  1. OS loads program and data into RAM (often by DMA).
  2. CPU fetches from RAM into registers and executes.
  3. Results go back to RAM.
  4. OS saves them to storage.
• Link: this is why adding RAM cures "thrashing" in virtual memory.

---

8. 📜 Past-Paper Questions & MCQs (CSS 2016–2026)

2017 · Paper-I · Q2(b) · 6 marks

Q. Describe the difference between Harvard and Von Neumann architectures of computers. Also discuss their traits in the light of their capabilities.

Quick revision + exam-ready answer

Quick revision — key points

• Von Neumann shares instruction/data path and address space
• Harvard separates instruction/data pathways
• Compare simultaneous access, cost and bottleneck
• Modern processors often use modified Harvard caches

Exam-ready answer — structured

Introduction / direct answer: Von Neumann: instructions and data occupy a common address space and commonly share the same memory path.

Main explanation

• It is flexible and easier to program, but competing instruction fetches and data transfers can produce the Von Neumann bottleneck. Harvard: instructions and data use separate memories and access paths; the processor can fetch an instruction and access data concurrently, improving predictable throughput.
• Harvard designs may require more hardware and complicate moving data between the two memory spaces. Application: classic Harvard systems are common in small embedded controllers and DSPs; general-purpose computers present a largely unified programming address space.
• Nuance: many modern processors use modified Harvard L1 instruction and data caches while maintaining a unified address space in main memory.

2026 · Paper-I · Q2(c) · 8 marks

Q. Your office plans to upgrade hardware but employees do not know which components matter. Which components would you explain to help them choose better systems?

Quick revision + exam-ready answer

Quick revision — key points

• Assess CPU and core performance for workload
• RAM capacity and SSD speed
• GPU, motherboard/ports and networking
• Power/cooling, upgrade compatibility and office scenarios

Exam-ready answer — structured

Introduction / direct answer: The best upgrade depends on the work performed, not simply the most expensive specification.

Main explanation

• CPU: assess generation, cores and responsiveness; more cores help sustained parallel workloads, while per-core performance matters for everyday apps. RAM: sufficient capacity prevents slowing caused by swapping; office multitasking typically benefits from adequate RAM and upgrade headroom.
• Storage: prefer an SSD for the operating system and applications; add capacity according to files and workloads. Motherboard/chipset: check compatibility, memory slots, expansion and suitable ports.
• GPU: integrated graphics is usually adequate for documents and video conferencing; dedicated GPUs make sense for 3D design, AI and demanding graphics. Display/input/networking: choose a clear ergonomic display, reliable peripherals, Wi-Fi/Ethernet and ports needed by staff.
• PSU/cooling: require stable power, safe wattage headroom and appropriate ventilation. Backup/security/UPS: include recovery and electrical protection in the decision.
• Recommended bundles: ordinary clerical office → modern midrange CPU, sufficient RAM and SSD; developer/analyst → stronger CPU, additional RAM and larger SSD; designer → powerful CPU/GPU, ample RAM and colour-accurate display. Conclusion: profile employees' tasks, inspect compatibility and prioritise upgrades that remove observed bottlenecks.