An Introduction for Computer Science and IT Students

---

Learning Objectives — By the end of this chapter, you should be able to:

1. Explain how information is represented and manipulated inside a computer (organization)
2. Describe how a computer's major subsystems are structured and interact (architecture)
3. Identify the physical components of a computer and their function (hardware)
4. Apply a systematic method to diagnose performance issues and hardware failures (support)

This chapter is organized around four pillars that, together, form a complete picture of "how a computer works":

• Organization (how data and instructions are represented and processed)
• Architecture (how subsystems are designed and connected)
• Hardware (the physical components themselves)
• Support (how to reason about and troubleshoot real systems).

Most introductory material only covers one or two of these — this article tries to give you all four, connected.

---

Part I — Computer Organization

Computer organization is concerned with how a computer actually carries out the instructions of the architecture — the operational, "under the hood" mechanics of representing and moving information.

1.1 What a Computer Is

A computer is a machine that accepts input, transforms it according to a stored set of instructions (a program), and produces output. Three properties distinguish it from ordinary electronics:

• Programmable — behavior is changed by loading new instructions, not by rewiring hardware.
• General-purpose — the same hardware can run a word processor, a browser, or a game.
• Automatic — once started, it executes a sequence of steps without a human pushing each one.

A microwave reacts to input, but it cannot be reprogrammed to do something categorically different. That distinction — programmability — is the foundation of everything else in this chapter.

1.2 Why Everything Is Binary

Digital circuits reliably distinguish exactly two states: current flowing or not flowing (often written 1 and 0). Trying to distinguish, say, ten separate voltage levels to represent decimal digits directly would leave very little margin for electrical noise, heat, or interference to cause a misread. Two states, with a wide safety margin between them, is what makes hardware fast, cheap, and dependable.

A single bit (binary digit) holds one of two values. Group 8 bits and you get a byte — 2⁸ = 256 possible values, enough to represent every character in the ASCII table. The byte became the universal unit of computer memory.

Representing numbers. Positive integers are written in binary using powers of two (1101 = 8+4+0+1 = 13). Negative integers use two's complement, a clever encoding that lets the same adder circuit handle both addition and subtraction. Numbers with a decimal point use floating-point representation (IEEE 754): a sign, a significand, and an exponent — essentially binary scientific notation, which is why some decimal values (like 0.1) can't be stored with perfect precision, the same way ⅓ has no exact decimal form.

Representing text, images, sound. Text maps characters to numbers via ASCII (128 values, English-only) or, more commonly today, Unicode (140,000+ characters covering virtually every writing system, typically encoded as UTF-8). An image is a grid of pixels, each storing red/green/blue intensity values. Video is a rapid sequence of images (frames) — compressed by storing only the differences between them. Sound is a continuous wave, digitized by sampling it thousands of times per second (44,100 times/sec for CD audio) and recording the amplitude at each instant.

💡 Key Insight — There is no special "text mode" or "image mode" inside a computer. Everything — a novel, a photograph, a symphony, a program — is the same kind of thing: a very long sequence of bits, interpreted differently depending on context.

1.3 The Instruction Cycle

Programs are broken down into machine instructions — small, precise operations the processor knows how to execute directly (e.g., "add these two numbers," "move this value"). A processor's complete vocabulary of such operations is its Instruction Set Architecture (ISA) — examples include x86-64 (most desktops/laptops) and ARM (most phones and increasingly laptops).

Every instruction is processed through the same three-stage cycle, repeated billions of times per second:

FETCH  →  DECODE  →  EXECUTE

• Fetch — retrieve the next instruction from memory, at the address held in a special register (the program counter).
• Decode — determine what operation the instruction represents.
• Execute — actually perform it, producing a result.

Two internal units make this possible: the Control Unit, which sequences and directs each stage (like a conductor), and the Arithmetic Logic Unit (ALU), which performs the actual math and logical comparisons. Registers — tiny storage locations built directly into the processor — hold the values being worked on at any given moment; because they sit inside the CPU itself, they are the fastest storage in the entire system.

---

Part II — Computer Architecture

Computer architecture is the higher-level blueprint: how major subsystems (processor, memory, storage, I/O) are organized and how they communicate to form a working system.

2.1 The Von Neumann Model

Proposed in the 1940s, the Von Neumann architecture remains the basic blueprint of nearly every computer today. Its defining idea: program instructions and the data they operate on live in the same memory, and the processor works through them sequentially.

        ┌─────────────┐        ┌──────────────┐
Input → │     CPU     │ ⇄ Bus ⇄ │  Memory (RAM) │ → Output
        │ (ALU + CU)  │        │ instructions   │
        └─────────────┘        │  + data        │
                                └──────────────┘

Before this model, reprogramming a machine often meant physically rewiring it — a process that could take days. The stored-program concept — instructions stored as data, in the same memory as everything else — is what made general-purpose, easily reprogrammable computers possible, and it directly enabled the modern idea of installable software.

2.2 The Von Neumann Bottleneck

Because instructions and data travel over the same shared pathway (the bus) between CPU and memory, and only one thing can move across it at a time, memory bandwidth becomes a limiting factor — even when the processor itself is capable of working much faster. The CPU frequently ends up waiting on memory. This has been a central challenge in computer engineering for decades, and much of modern hardware design exists specifically to work around it:

2.3 The Memory Hierarchy

No single memory technology is simultaneously fast, large, and cheap — so architecture uses a hierarchy, trading capacity for speed at each level:

Registers  →   Cache (L1→L2→L3)   →   RAM   →   Storage (SSD/HDD)
 fastest         fast, small          medium      slow, huge, permanent
 ~picoseconds    ~nanoseconds        ~100ns      ~0.1–10 milliseconds

• Registers sit inside the CPU — near-instant, but there are only a handful.
• Cache (L1/L2 per-core, L3 typically shared) buffers the gap between CPU and RAM.
• RAM holds whatever program and data are actively in use — fast, but volatile: its contents vanish the instant power is lost.
• Storage (SSD/HDD) is slow by comparison but non-volatile, retaining data indefinitely without power.

When RAM fills up, the operating system can borrow space from storage as virtual memory, via a swap file — but because storage is orders of magnitude slower than RAM, heavy swapping causes the sharp, characteristic slowdown of a memory-starved system.

2.4 Parallelism: Cores, Threads, and the GPU

A core is a complete, independent processing unit; a modern CPU typically has 4–16+ cores. A thread is a software-level stream of instructions; simultaneous multithreading (SMT/Hyper-Threading) lets one physical core interleave two threads to use its internal resources more fully — it is not the same as having two real cores.

CPUs are optimized for a small number of complex, varied tasks executed quickly in sequence. GPUs invert this trade-off entirely: thousands of simpler cores, each less capable individually, but built to execute the same operation across massive amounts of data simultaneously (parallel computing). This is ideal for rendering millions of pixels, and — because training a neural network is fundamentally repeated matrix multiplication — it is equally ideal for AI workloads, which is why GPUs (often with dedicated Tensor Cores) dominate machine learning today.

---

Part III — Hardware

Hardware is where organization and architecture become physical, touchable components.

3.1 The Motherboard

The motherboard is the physical backbone that every other component connects to, both electrically and for communication.

Components communicate over buses: a data bus (the payload), an address bus (the destination), and a control bus (the signaling — read, write, timing). Compatibility hinges on matching socket type, supported memory generation (DDR4/DDR5), physical form factor (ATX, Micro-ATX, Mini-ITX), and the chipset's feature set.

3.2 Storage Devices

Data is organized in layers: disk → partitions → file system. A partition table (modern GPT, or legacy MBR) records how a disk is divided; a file system (NTFS, ext4, APFS...) then organizes actual files and folders within each partition. The TRIM command helps SSDs maintain performance over time by proactively erasing blocks freed by deleted files.

3.3 Power and Cooling

The Power Supply Unit (PSU) converts high-voltage alternating current from the wall into stable, low-voltage direct current for internal components.

Power (W) = Voltage (V) x Current (A)

Efficiency matters: a PSU's 80 PLUS rating (Bronze through Titanium) certifies how much input power is actually delivered to components versus lost as heat. A poor-quality PSU can deliver unstable voltage that damages sensitive components over time.

Every electrical component generates heat as a byproduct of resistance. Conduction transfers that heat into a metal heatsink (with thermal paste filling microscopic surface gaps for better contact); convection, driven by fans or liquid coolers, carries it away into moving air or liquid. If temperature rises too far, the processor engages thermal throttling — deliberately reducing its own clock speed to avoid damage, at the cost of performance. Good case airflow (intake at the front/bottom, exhaust at the back/top) matters as much as any individual heatsink.

3.4 Networking Hardware

A Network Interface Card (NIC) connects a computer to a network, wired (Ethernet, via RJ45 connectors — stable, fast) or wireless (Wi-Fi — convenient, more prone to interference).

Every device carries two distinct addresses:

A switch connects devices within one local network using MAC addresses; a router connects separate networks (e.g., a home network and the Internet) using IP addresses; a modem translates the ISP's incoming signal into a usable digital one. Home routers usually combine all three roles into a single box.

---

Part IV — IT Support: Diagnostics and Systems Thinking

Knowing what components do is only half the picture; the other half is reasoning about how they behave together — and knowing what to do when something breaks.

4.1 From Power Button to Desktop: The Boot Sequence

⚡ Power applied
     ↓
🔌 PSU converts current
     ↓
🌱 Firmware (BIOS/UEFI) runs — before any OS exists
     ↓
✅ POST (Power-On Self-Test) verifies core hardware
     ↓
🔧 Firmware initializes RAM, storage controllers, etc.
     ↓
🥾 Bootloader is located (per the configured boot order) and loaded
     ↓
🧬 Bootloader loads the OS kernel into RAM and hands off control
     ↓
🖥️ Kernel initializes drivers/services → GPU renders the desktop

Modern systems add security layers to this chain: Secure Boot verifies, via digital signatures, that the bootloader hasn't been tampered with, and a TPM (Trusted Platform Module) securely stores encryption keys used for disk encryption and system integrity checks.

4.2 Systems Thinking: The Bottleneck Concept

No component operates in isolation — a fast CPU sitting idle, waiting on a slow disk, delivers exactly the disk's performance, not the CPU's. The bottleneck is whichever component is saturated (near 100% utilization) while others have headroom; it — and only it — sets the pace for the whole system. This has a direct practical consequence: upgrading a component that isn't the bottleneck yields little to no improvement. Identifying the actual bottleneck (via monitoring tools tracking CPU%, RAM/swap usage, disk I/O, temperature, and GPU%) should always precede any upgrade decision.

4.3 A Structured Diagnostic Method

Effective troubleshooting follows a repeatable process rather than guesswork:

1. Observe the symptom precisely.
2. List plausible causes.
3. Prioritize the most likely one — starting with the simplest, most common causes first.
4. Test that specific hypothesis.
5. Interpret the result — confirmed or ruled out?
6. Resolve, or return to step 3 with the next hypothesis.

⚠️ Common Pitfall — Assuming every slowdown is a hardware failure. A great deal of "the computer is slow" turns out to be software-level: too many background processes, a near-full disk, or an operating system that hasn't been maintained — not a broken part.

---

Chapter Summary

A computer is best understood as four layers built on top of each other:

HARDWARE (physical components)
     ↓ organized according to
ARCHITECTURE (how subsystems are designed and connected)
     ↓ operating through
ORGANIZATION (how data and instructions are actually processed)
     ↓ kept running via
SUPPORT (diagnosing, maintaining, and reasoning about the whole system)

No layer is complete without the others. Knowing that a GPU has thousands of cores (hardware) means little without understanding why parallelism helps certain workloads (architecture) and how those cores actually execute instructions (organization) — and none of that helps you fix a slow machine without a disciplined diagnostic method (support). Together, these four perspectives are what separate someone who can name computer parts from someone who can actually reason about a computer as a system.

---

Key Terms

bit · byte · ISA · fetch-decode-execute · Von Neumann architecture · bottleneck (Von Neumann) · cache · pipelining · memory hierarchy · volatile / non-volatile · virtual memory / swap · core / thread · parallel computing · motherboard / chipset / bus · HDD / SSD / NVMe · partition / file system · PSU / 80 PLUS · thermal throttling · MAC / IP address · BIOS/UEFI / POST / bootloader / kernel · bottleneck (systems)

Review Questions

1. Why did computer designers settle on binary rather than a system with more than two states?
2. Explain, in your own words, the Von Neumann bottleneck — and name two hardware techniques that mitigate it.
3. Why is RAM described as volatile and storage as non-volatile? What practical consequence does this have?
4. A computer has a very fast CPU but feels sluggish during everyday use. List three possible causes and how you would test each one.
5. Explain why GPUs are well suited to both graphics rendering and AI training, using the concept of parallelism.
6. Walk through the boot sequence from power-on to desktop, naming the role of each stage.