Overview#

Understanding computer hardware is the foundation of IT literacy. Every piece of software ultimately runs on physical components — processors, memory, storage, and interconnects. Knowing how these parts work, interact, and fail gives you a mental model for performance tuning, troubleshooting, and system design.

The Central Processing Unit (CPU)#

The CPU executes instructions. It’s the brain of the computer, but a very specific kind of brain — one that does billions of simple operations per second.

Architecture#

flowchart TB
    subgraph CPU["CPU Package"]
        subgraph Core1["Core 0"]
            ALU1["ALU"]
            FPU1["FPU"]
            L1C1["L1 Cache (64KB)"]
        end
        subgraph Core2["Core 1"]
            ALU2["ALU"]
            FPU2["FPU"]
            L1C2["L1 Cache (64KB)"]
        end
        L2["L2 Cache (per core, 256KB-1MB)"]
        L3["L3 Cache (shared, 8-64MB)"]
        MC["Memory Controller"]
        PCIE["PCIe Controller"]
    end
    MC --> RAM["System RAM"]
    PCIE --> GPU["GPU / NVMe / Devices"]

Key Concepts#

The Instruction Pipeline#

Modern CPUs don’t execute one instruction at a time. They use pipelining — breaking execution into stages so multiple instructions are in-flight simultaneously:

Fetch → Decode → Execute → Memory Access → Write Back

CPU Families#

x86 vs ARM#

Memory (RAM)#

RAM is volatile, fast storage that the CPU accesses directly. When the system loses power, RAM contents are lost.

Memory Hierarchy#

Speed and capacity are inversely related:

graph TB
    A["CPU Registers<br/>~1 cycle, bytes"] --> B["L1 Cache<br/>~4 cycles, 64KB"]
    B --> C["L2 Cache<br/>~12 cycles, 256KB-1MB"]
    C --> D["L3 Cache<br/>~40 cycles, 8-64MB"]
    D --> E["RAM (DRAM)<br/>~100 cycles, 8-128GB"]
    E --> F["SSD/NVMe<br/>~10,000 cycles, 256GB-4TB"]
    F --> G["HDD<br/>~10,000,000 cycles, 1-20TB"]

DDR Generations#

Key RAM Concepts#

• Channels — Dual-channel doubles bandwidth by using two sticks in parallel. Always install RAM in pairs for matched channels.
• ECC (Error-Correcting Code) — Detects and corrects single-bit errors. Required for servers (Xeon/EPYC), not supported on consumer platforms.
• CAS Latency (CL) — The delay between requesting data and receiving it, measured in clock cycles. Lower is better, but raw speed often matters more.
• DIMM vs SO-DIMM — Desktop modules vs laptop modules (smaller form factor).

How Much RAM Do You Need?#

Storage#

Storage retains data without power. The two main technologies are fundamentally different:

HDD (Hard Disk Drive)#

Mechanical: spinning magnetic platters with a read/write head on an actuator arm.

SSD (Solid State Drive)#

Electronic: NAND flash memory with no moving parts.

NAND Flash Types#

Form Factors#

Graphics Processing Unit (GPU)#

GPUs are massively parallel processors. Originally for rendering graphics, now critical for machine learning, scientific computing, and video processing.

CPU vs GPU Architecture#

GPU Memory (VRAM)#

Use Cases by Domain#

The Motherboard#

The motherboard is the central PCB connecting all components.

Key Components#

flowchart TB
    CPU["CPU Socket"] --- Chipset
    Chipset --- RAM["RAM Slots (DIMM)"]
    Chipset --- PCIE["PCIe Slots (x16, x4, x1)"]
    Chipset --- M2["M.2 Slots (NVMe/SATA)"]
    Chipset --- SATA["SATA Ports"]
    Chipset --- USB["USB Headers"]
    Chipset --- NET["Ethernet / WiFi"]
    Chipset --- AUDIO["Audio Codec"]
    CPU --- RAM
    CPU --- PCIE

Socket Types#

Chipset Role#

The chipset manages communication between the CPU and lower-speed peripherals (SATA, USB, audio, extra PCIe lanes). Higher-tier chipsets provide more:

• PCIe lanes
• USB ports
• Overclocking support
• Multi-GPU support

Form Factors#

Power Supply Unit (PSU)#

The PSU converts AC mains power to the DC voltages components need (+12V, +5V, +3.3V).

Key Specifications#

How Much Wattage?#

Rule of thumb: Add up CPU TDP + GPU TDP + 100W for the rest, then add 20–30% headroom.

ATX Power Connectors#

Buses and Interconnects#

Buses are communication pathways between components.

PCIe (Peripheral Component Interconnect Express)#

The primary high-speed bus for GPUs, NVMe SSDs, network cards, and more.

Lane widths: x1, x4, x8, x16. A device can use fewer lanes than the slot provides.

Other Interconnects#

Cooling#

Every watt of power consumed becomes heat. Without cooling, components throttle or fail.

Cooling Methods#

Thermal Interface Material (TIM)#

Thermal paste fills microscopic gaps between the CPU die and heatsink. Quality matters — a good paste can mean 5–10°C difference under load.

Airflow Principles#

• Positive pressure (more intake than exhaust) — keeps dust out, preferred for most builds
• Negative pressure (more exhaust than intake) — slightly better cooling, pulls in dust from cracks
• Front/bottom intake, top/rear exhaust — follows natural convection (hot air rises)

Peripherals and I/O#

Input Devices#

• Keyboard (mechanical, membrane, capacitive)
• Mouse (optical, laser)
• Touchscreen (capacitive, resistive)
• Stylus/digitizer (Wacom, Apple Pencil)
• Microphone, webcam, scanner

Output Devices#

• Monitor (IPS, VA, OLED — resolution, refresh rate, response time)
• Speakers / headphones
• Printer (inkjet, laser)

Expansion Cards#

• Dedicated GPU
• Network card (10GbE, WiFi 6E/7)
• Sound card
• Capture card
• RAID controller

Server vs Desktop vs Laptop Hardware#

Key Takeaways#

1. CPU performance is determined by cores × IPC × clock speed — not just GHz
2. Memory hierarchy matters — cache misses are expensive (100x slower to reach RAM vs L1)
3. Storage interfaces matter as much as the drive — NVMe over PCIe is ~10x faster than SATA
4. GPUs are not just for gaming — they’re the backbone of modern AI and parallel computing
5. Power and cooling are engineering constraints — a 450W GPU needs appropriate PSU and airflow
6. Form factor choices cascade — pick the motherboard size first, then the case, then cooling constraints follow
7. Server hardware prioritizes reliability and remote management over raw consumer performance