ChipFACTORYFoundations
Apply now
Foundations/STAGE 00
Foundations Track/STAGE 00
15 min first-principles journey
SECTION 0

What is this strange thing called a chip?

The Big Question

“We use chips everywhere, but what exactly is a chip, and how can something so small do so much?”

Welcome to the beginning of your engineering journey. Rather than reading abstract definitions, we will open a smartphone, peel back the black protective package, and explore how physical silicon switches perform computation.

0.1The Physical Path of an Action

Where does computation actually happen?

Every second, billions of humans tap glass screens on smartphones. But what executes the intent behind that touch? Is software itself doing the work?

What happens when I open an app on my phone?

When you tap an icon, software initiates a digital request. But software is just a static plan—like sheet music or a recipe book. It contains the instructions, but it cannot bake the cake or produce sound on its own.

Where does the calculation actually happen?

The operating system translates your tap into low-level machine instructions and passes them to physical hardware (CPU, GPU, RAM) where electric currents physically toggle switches to compute output.

Is software itself doing the calculation?

No. Software is only the instruction set. Physical electrons moving through physical conductive pathways perform every calculation.

Pedagogy: Slide 1 & 3Interactive Discovery

How an App Works: The Chain of Command

When you tap an app, software alone cannot perform work. Follow the request down to the physical silicon engine.

Step 1: Choose User Action
1. SOFTWARE / APP

Customer's Order

Tapping an icon creates an intent in the user interface.

Triggered Request:
Customer: 'Multiply 849 by 372'
2. OPERATING SYSTEM

The Waiter / Translator

Translates the human request into raw binary machine instructions.

Decoded System Call:
OS: Load MUL instruction into register R1 & R2, pass to CPU scheduler thread
3. PHYSICAL HARDWARE

Kitchen Staff / Silicon

Transistors physically toggle electrical current to calculate and produce output.

Physical Execution:
Hardware: CPU Arithmetic Logic Unit (ALU) switches transistor logic gates to yield 315,828
Active Hardware:CPU ALU + Registers
Approved Teaching Analogies
Software = The Recipe Book

A recipe contains every exact step to bake a cake, but a piece of paper cannot heat an oven or mix batter. On its own, software is static, inert text.

Hardware = The Kitchen & Chefs

The chefs, pans, heat, and raw ingredients physically execute the recipe to produce a meal. The processor is the chef, RAM is the prep table, and Flash is the pantry.

Software is just a static plan (sheet music / recipe). Without physical hardware to switch electrical currents, nothing happens.

What part of the phone executes the instructions?

The processor. A modern smartphone houses multiple specialized processors inside a single System on Chip (SoC): the CPU handles general logic and sequential tasks, while the GPU deploys thousands of shader cores in parallel for visuals.

Pedagogy: Slide 4Interactive Discovery

Instruction Execution: CPU vs. GPU

Inside a modern System-on-Chip (SoC), different silicon engines specialize in different computational physics.

Choose Test Workload:
64 independent math operations

Central Processing Unit (CPU)

The General Engine • 2-8 Big Cores @ 3.4 GHz
0%

Optimized for low-latency, complex control flow, memory paging, and sequential step-by-step algorithms.

Core 0 & 1 Serial Pipeline
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Key Trait: Slow on 64 items (processes serial batches one after another).

Graphics Processing Unit (GPU)

The Parallel Engine • 1000s of Shader Units
0%

Optimized for SIMD (Single Instruction, Multiple Data) workloads: images, 3D vertices, video decoding, AI tensors.

SIMD Parallel Matrix Array (64 Cores Active)
Key Trait: Instant blitz (all 64 pixels computed simultaneously in parallel).
A CPU is like a high-speed sports car for complex individual routes. A GPU is like a massive 1,000-seat train moving everyone at once.

What is a processor?

A processor transforms static software code into active physical computation by continuously cycling through three hardware stages: Fetch (pulls instruction from memory), Decode (interprets opcode), and Execute (toggles transistor switches).

Pedagogy: Slide 5Interactive Discovery

The Processor: Central Engine

The processor executes trillions of instructions by endlessly repeating three fundamental hardware stages: Fetch, Decode, and Execute.

Clock Speed:
Cycles: 0
STAGE 01

Fetch

Continuously retrieves the next required instruction directly from memory (RAM / Cache).

Program Counter:
0x0004
Fetched Opcode:
LOAD R1, [0x100]
STAGE 02

Decode

Interprets incoming instructions to determine the exact logical or mathematical operations needed.

Instruction Type:
Memory Read
Control Signals:
Opcode: LOAD • Target Register: R1 • Source Address: RAM 0x100
STAGE 03

Execute

Performs arithmetic and logical actions via physical transistors acting as rapid electrical switches.

ALU Work:
Memory bus pulls value 42 from RAM cache line into R1 register latch
Register Output:
R1 = 42
CPU Internal Architecture & Data Path
Active Stage: Stage 01
RAM / Cache
Instruction Bus
Instruction Reg
IR: LOAD
Control Unit
Instruction Decoder
ALU / Transistors
R1 = 42
Modern phone CPUs repeat this loop over 3,000,000,000 times every second (3+ GHz), powered by microscopic transistor switches.

What is a memory?

Memory holds bits (1s and 0s). Computers require two distinct forms of memory: a fast temporary workbench (RAM) for active tasks that clears when powered off, and a permanent bookshelf (Flash storage) that retains files long-term.

Pedagogy: Slide 6Interactive Discovery

Understanding Memory: The Workspace Analogy

Memory holds bits (1s and 0s). Understand why computers need both a fast temporary workbench (RAM) and a permanent bookshelf (Flash).

⚡ Power rail active (1.8V / 1.1V)
Temporary Workbench

Random Access Memory (RAM)

Volatile Active

Acts as the fast, temporary workbench or scratchpad where the processor executes active tasks.

Active Workbench Items (3)Latency: ~10 ns
Active OS Kernel Thread
2.1 GB
Instagram Live Feed Buffer
450 MB
Current Math Expression: 849 × 372
64 KB
• Access: Instantly accessed by CPU registers (~10 ns)
• Purpose: Active app state and variables
• Persistence: Temporary (lost when powered off)
Long-Term Bookshelf

Flash Storage (NAND / UFS)

Non-Volatile (Permanent)

Acts as the library or bookshelf for long-term retention of all photos, apps, and operating system files.

Stored Library Items (4)Latency: ~50,000 ns
Operating System (Android / iOS)
14 GB
IMG_9482_Goa_Vacation.jpg
4.2 MB
Instagram App Bundle
220 MB
Calculator App
12 MB
• Access: Slower; must be retrieved into RAM before CPU can work on it
• Purpose: OS image, photos, videos, downloaded apps
• Persistence: Permanent (trapped electron charge lasts decades)
RAM is extremely fast (nanoseconds) but volatile (loses charge without power). Flash is permanent but slower to read/write.

What other chips exist inside a phone?

A phone is not just one chip—it is an entire ecosystem of specialized silicon dies. High voltages, radio frequencies, memory cells, and display drivers each require distinct semiconductor manufacturing processes.

Pedagogy: Slide 7Interactive Discovery

Smartphone Motherboard Teardown: A Complex City of Chips

A modern phone is not just one chip—it is an entire ecosystem of specialized silicon dies working in synchronization.

PCB Teardown Map (Click any chip)High-Density Interconnect PCB
BATTSoC BRAINThe BrainRAMHigh-Speed MemoryPMICPower GridRFWireless CommunicationsSTORAGENon-Volatile StorageDISPLAYDisplay & Timing
💡 Tip: Click each IC package to inspect why specialized silicon is required.
The Brain

System on Chip (SoC)

Primary Function:

Houses CPU, GPU, NPU, and ISP. Executes all high-level software instructions, graphics, and AI models.

Why Can't This Be In The Main CPU?

Manufactured on leading-edge 3nm/4nm process for maximum transistor density (~15 billion transistors).

Engineering Specs:
3.4 GHz Octa-Core • Adreno/Immortalis GPU • 16-Core NPU
Quick Select Component:
Different chips use completely different semiconductor physics (e.g., 3nm logic vs high-voltage power vs 3D NAND storage).
Subsection 0.1 Complete
NEXT: 0.2 WHAT EXACTLY IS A CHIP?
0.2The Physical Anatomy

What exactly is a chip?

When you hold an integrated circuit, you see a flat black square with shiny metal pins. But is that black rectangle the chip?

What does a chip physically look like?

From the outside, it looks like a small black plastic or ceramic square with metal legs or solder balls on the bottom.

Is the chip the black package we see?

No. The black rectangle is merely protective packaging—functioning like a helmet or suit of armor to shield the fragile heart inside.

What is actually inside the package?

Hidden deep inside is a paper-thin square of polished silicon called a die, often smaller than a human fingernail, connected to external pins by microscopic gold bond wires (ø18–25μm).

Pedagogy: Slide 8, 9 & 10Interactive Discovery

Anatomy of an Integrated Circuit: The Black Square is Armor

What you see on a circuit board is not the chip. It is a protective helmet of black epoxy. Drag the slider to decapp the package.

X-Ray / Decapping Cutaway Layer:50% Cross-Section Cutaway
External ArmorBare Die
IC Package Cross-Section ModelDual In-line / QFP Package
SILICON DIEACTIVE SILICON DIE (~0.3mm THICK)GOLD BOND WIRES (ø18-25μm)EXTERNAL COPPER PINS
Packaging Principle

The Helmet Analogy

1. Protective Epoxy Molding:

The black plastic package is like a soldier's helmet or suit of armor. It shields the fragile silicon crystal from humidity, atmospheric oxygen, dust, and mechanical vibration.

2. The Silicon Die:

Hidden deep inside is a paper-thin square of pure silicon. All calculation, memory storage, and logic happen exclusively on this microscopic shiny die.

3. Micro Bond Wires / Micro-Bumps:

Delicate gold wires—thinner than a human hair (ø18-25μm)—bridge the die bond pads to the sturdy external pins soldered to the motherboard.

💡 Takeaway: Never judge a chip by its package dimensions. A giant black IC package might contain a die no larger than a grain of rice.
The true active chip is the tiny, paper-thin square of polished silicon die at the center, smaller than a fingernail.

What is silicon?

Silicon is the 2nd most abundant element on Earth, found naturally in quartz beach sand (SiO₂). Raw sand is purified at 1425°C into hyper-pure single-crystal ingots (99.9999999% purity) and sliced into circular wafers.

Pedagogy: Slide 11Interactive Discovery

Silicon: From Beach Sand to Microchips

How does ordinary quartz beach sand become the flawless atomic canvas for 15 billion microscopic transistors?

🏖️
Raw Mineral State

Beach Sand (SiO₂)

Abundant, impure, non-conductive quartz grains

Chemical & Physical Transformation

Beach Sand (Quartz)

Material State:SiO₂ (Silicon Dioxide)
Purity Grade:~95% Mineral Purity
Process Temperature:Ambient (25°C)

Silicon is the 2nd most abundant element in Earth's crust. It exists naturally as ordinary beach sand bound tightly with oxygen atoms.

Electronic-grade silicon requires 99.9999999% purity ('nine nines')—fewer than one impurity atom per billion silicon atoms.

Why is silicon used?

Conductors (copper) are always ON; insulators (rubber/glass) are always OFF. Silicon is a semiconductor: introducing precise chemical dopants allows engineers to dynamically control electron flow via applied gate voltages.

Pedagogy: Slide 12Interactive Discovery

Silicon: The Foundational Semiconductor

Why is silicon the foundation of all computing? Because it is neither a pure conductor nor a pure insulator—it is an electrically controllable switch.

Applied Gate Terminal Voltage (Vgs)Toggling voltage dynamically changes silicon from insulator to conductor
Material Channel:semiconductor
Conduction State:⚡ CURRENT FLOWING (LOGIC 1)
SOURCE
DRAIN
e⁻
e⁻
e⁻
e⁻
e⁻
e⁻
e⁻
e⁻
Energy Band Gap:Moderate gap (1.1 eV). Applied voltage easily creates a temporary conductive path!
Why Useful For Chips:The building block of transistors! Allows binary switching (0 and 1).
Binary Logic Mapping:High Voltage State = '1'
Dynamically switching electrical current ON and OFF at will is the foundational physical property of modern computing.

What is a transistor?

A transistor is an electrically operated switch with no physical moving parts. Applying a voltage to the Gate terminal creates an electrostatic field that opens a temporary conductive channel between Source and Drain.

Pedagogy: Slide 13Interactive Discovery

How a Transistor Operates: The Voltage-Controlled Switch

A transistor has no mechanical moving parts. It uses electric voltage at the Gate to open or close an electron channel in silicon.

Gate Terminal Voltage (Vgs): 1.20 Volts
Switch State:SWITCH ON (LOGIC 1)
0.0V (Closed)1.8V (Fully Open)
Threshold Voltage (Vth) = 0.60V
NMOS Transistor Cross-Section (40nm Scale)MOSFET Physics Diagram
P-TYPE SILICON SUBSTRATE (HOLES / NO FREE ELECTRONS)SOURCE (N+)DRAIN (N+)SiO₂ GATE OXIDE (INSULATOR)GATE (Vgs = 1.20V)SOURCE (S)DRAIN (D)⚡ CONDUCTIVE N-CHANNEL FORMED→→→→
Transistor Mechanism

Using Electricity to Control Electricity

1. When Gate = 0.0V (Off State):

The P-type silicon substrate acts as a physical barrier. No electrons can cross between the Source and Drain wells. Circuit is OPEN (Binary 0).

2. When Gate ≥ 0.6V (On State):

Positive charge on the gate attracts free electrons to the surface, creating an inversion layer (the N-Channel). Current freely flows across. Circuit is CLOSED (Binary 1).

Zero Moving Parts: Because it switches via pure electrostatics (electric field), it can toggle billions of times every single second without mechanical wear!
Applied Gate voltage dynamically forms the temporary conductive N-Channel within the P-type silicon substrate.

Why do we need billions of transistors?

A single transistor handles only one binary choice (1 or 0). Combined into dense arrays of billions switching at gigahertz speeds, they execute 3D graphics, artificial intelligence, and operating systems simultaneously.

Pedagogy: Slide 14Interactive Discovery

Scaling Computational Power: Micro to Massive

A single transistor can only answer 'yes' or 'no'. How do billions of them working together create modern intelligence?

Select Transistor Scale Hierarchy:15 Billion Transistors (Modern Smartphone SoC (3nm))
📱

15 Billion Transistors

Modern Smartphone SoC (3nm)
Transistor Density MapClock: 3.4 GHz - 4.3 GHz
~15,000,000,000 Switches on 3nm Silicon
Emergent Capability

What Can Be Computed At This Scale?

Real-time 3D ray-traced graphics @ 120 FPS
Generative AI neural network inference on-device
4K 60fps HDR video capture & spatial audio
Multi-gigabit 5G wireless carrier aggregation
Benchmark Reference:
Apple A17 Pro / Snapdragon 8 Elite (15-20 Billion Transistors)
By assembling billions of microscopic switches onto one silicon die, we graduate from a single bit flip to real-time AI and 3D graphics.
0.3From Physical Reality to Logic

How can something physical perform computation?

How does an inert piece of silicon and copper represent numbers, make decisions, and execute programs?

How can a piece of silicon represent information?

Data in hardware is physical voltage, not ink or text. Silicon dies contain millions of microscopic lattice points holding electric charge: HIGH voltage (1.8V) represents binary '1', and LOW voltage (0V) represents binary '0'.

Pedagogy: Slide 15Interactive Discovery

Representing Digital Data in Silicon

Data in hardware is physical voltage, not ink, text, or pixels. Click nodes in the 2D silicon lattice to store electric charge.

Load Pattern:
1.8V (HIGH / '1'): 12
0.0V (LOW / '0'): 84
1
2
3
4
5
6
7
8
9
10
11
12
A
B
C
D
E
F
G
H
Physical Representation

Microscopic Voltage States

Information is never stored in silicon with ink or text. Instead, data is encoded entirely by whether electrical charge is trapped at specific lattice coordinates.

Charge at Lattice Point:1.8 Volts vs 0.0 Volts
Physical State:Inversion Charge
💡 By controlling where voltage is HIGH vs LOW across billions of grid points, chips encode, store, and process digital photos, music, and software programs purely as physical charge.
HIGH VOLTAGE (1.8V) = Binary '1' • LOW VOLTAGE (0V) = Binary '0'.

How can electricity represent a number?

By grouping multiple physical copper wires side-by-side into a 'bus'. Each wire carries a power-of-two mathematical weight (2⁰, 2¹, 2², 2³). Wire voltages directly form binary numbers that adders calculate.

How can electricity represent a decision?

By routing voltages through logic gates (AND, OR, NOT). If two voltage inputs are HIGH, the gate outputs a HIGH signal—physically executing a logical 'decision' without human thought.

Pedagogy: Slide 16Interactive Discovery

Representing Numbers via Voltage & Bus Wires

How does electricity represent a number? By grouping physical parallel copper wires into a 'bus' and driving discrete voltages on each wire.

Voltage Encoding:HIGH (1.8V) = '1'LOW (0.0V / GND) = '0'
4-Wire Parallel Copper BusClick switch to toggle voltage
Wire 4
2³ (Weight: 8)
0.0V (GND / LOW)
0
Wire 3
2² (Weight: 4)
⚡ 1.8V (HIGH)
1
Wire 2
2¹ (Weight: 2)
0.0V (GND / LOW)
0
Wire 1
2⁰ (Weight: 1)
⚡ 1.8V (HIGH)
1
Live Physical Bus Reading

Decoded Number

Binary Representation (4 Bits):
0101₂
= Decimal 5
Calculated Formula:
(0 × 2³) + (1 × 2²) + (0 × 2¹) + (1 × 2⁰)
= 0 + 4 + 0 + 1 = 5
💡 A 64-bit computer bus works on the exact same physical principle—just 64 parallel copper traces carrying simultaneous voltages!
Physical wire voltage combinations directly constitute complete binary digital numbers.

How can billions of tiny devices work together?

A high-speed global clock signal acts like a master conductor's baton. It ensures all microscopic switches transition states simultaneously on rising clock edges, preventing data collisions across the hierarchy.

Pedagogy: Slide 18Interactive Discovery

Hardware Hierarchy & Synchronization

How do billions of tiny switches work together without chaotic collisions? A global clock signal coordinates every transistor like a master conductor's baton.

Clock Level:LOW (0.0V)Beats: 0
Clocked Synchronous PipelineD Flip-Flop Registers
Master Clock Line (CLK)Waiting for next edge
Register A (In 1)
12
Latched @ CLK
Register B (In 2)
30
Latched @ CLK
Output Register
42
A + B = Result
Structural Hierarchy

How Microscopic Scales Combine

1. Transistors:Microscopic physical switches toggling electrical pulses.
2. Logic Gates (NAND, OR, NOT):Pairs of transistors executing elementary boolean logic.
3. Functional Blocks:Adders, multipliers, registers, and multiplexers.
4. Complete Processing Engines:Full multi-core CPU & GPU engines operating in clock unison.
On each rising clock edge (↑), flip-flop registers latch new data simultaneously across the entire chip hierarchy.

Is a chip programmed like software?

Software is virtual traffic navigating road directions. Hardware is the concrete city itself—permanent physical pathways etched into silicon.

Pedagogy: Slide 19Interactive Discovery

Is a Chip Programmed Like Software?

Software is virtual directions given to a machine. Hardware is the permanent, immutable physical city of roads and bridges.

Software: Ephemeral Traffic

Flexible & Malleable

• Virtual & Sequential: Software consists of text instructions stored temporarily in RAM that a processor fetches and executes line by line.

• Instant Modification: If you find a bug, you simply rewrite a line of code and push an update in seconds without touching physical silicon.

“Like GPS route directions telling cars which turn to take at an intersection.”

Hardware: Concrete Infrastructure

Permanent Physical Pathways

• Immutable Silicon: Chip design physically etches billions of microscopic copper wires and transistors into silicon wafers.

• Zero Post-Tapeout Edits: Once manufactured at a foundry (TSMC, Intel), you cannot 'update' a physical transistor wire. It is set in stone forever.

“Architecting the physical city itself—pouring concrete for bridges, tunnels, and highways.”
You don't 'write' a chip like software; you physically construct permanent silicon pathways for electricity to travel upon.

Or is the hardware itself designed to perform a particular function?

General-purpose CPUs are flexible Swiss Army Knives that can execute any code, while specialized ASICs are hardwired express highways permanently designed for extreme speed on a single task.

Pedagogy: Slide 20Interactive Discovery

Physical Hardware Dictates Function: CPU vs. ASIC

Does a chip adapt to software, or is the silicon physically hardwired for a single specialized purpose?

⚡
Dedicated Highway Architecture

Application-Specific IC (ASIC)

Zero software decoding overhead. Logic gates are physically wired directly into the mathematical equation.

Raw Stream══════ DIRECT SILICON HARDWIRED PIPELINE ══════Instant Out
Engineering Trade-Offs

ASIC Trade-Offs

Speed & Efficiency: 100x–1000xComputes tasks (like AI matrix multiplication or H.265 video encoding) with extreme speed and minimal battery drain.
Flexibility: 0%Cannot run a web browser or operating system. Permanently wired for its one dedicated job.
💡 In a smartphone SoC, the CPU handles general app logic while specialized ASICs (like the NPU, ISP, and Video Decoder) handle heavy compute tasks!
A CPU is a Swiss Army Knife that can do anything slowly. An ASIC is a dedicated rocket engine permanently wired for one task.
0.4The Journey to Tapeout

If we wanted to build our own chip, where would we start?

How do we decompose a machine as complex as a modern smartphone into simple, understandable engineering primitives that you can build yourself?

How do we break down a massive system into simple primitives?

By deconstructing the scale through six continuous layers of abstraction: from the smartphone glass down through the SoC, functional blocks, logic gates, and individual transistors to the raw silicon lattice.

What are the layers of abstraction between software and silicon?

Explore the camera zoom depth below to travel from the macroscopic consumer device down to individual atoms.

Pedagogy: Slide 2 & Section 0.4Interactive Discovery

The 6-Level Zoom Hierarchy: Smartphone to Silicon

Chip design is the art of navigating six orders of magnitude—from a handheld phone down to individual silicon atoms.

Current Camera Depth:Level 1 of 6: Modern Smartphone
Scale: ~150 mm (15 cm)
Physical Scale: ~150 mm (15 cm)

Modern Smartphone

Consumer Hardware Ecosystem
Abstraction Level 1

Modern Smartphone

The familiar physical device holding the screen, battery, antenna, cameras, and motherboard.

Internal Physical Components:
OLED Glass Display
Li-Ion Battery
Motherboard PCB
Camera Sensor Modules
Calculation happens inside microscopic physical switches on silicon. Billions of electrical pulses turn on and off, proving that complex processing relies on simple physical building blocks.

How does a beginner go from logic to a manufactured tapeout?

By following the 6-stage chip design flow: starting with high-level specifications, writing RTL in Verilog, verifying in simulation, synthesizing with OpenROAD into physical silicon geometries, and submitting for tapeout fabrication.

Pedagogy: Slide 21Interactive Discovery

Chip Design Flow: Specification to Silicon

How does an idea transform into manufactured silicon? Follow the 6-stage engineering design methodology.

STAGE 01 • SPECOpen-Source EDA Toolchain

Problem Statement & Specs

Output:PPA Goals (Performance, Power, Area)

Defines the exact problem the chip must solve, establishing clock frequency targets (e.g. 100 MHz vs 3 GHz), power budgets (e.g. 5 Watts), and maximum die area (e.g. 10mm²).

Primary Tools:System Architecture Docs
Course Coverage:Covered In Architecture Overview
In this course, you will learn to write Verilog RTL (Stage 03), verify in simulation (Stage 04), and synthesize with OpenROAD (Stage 05)!
Section 0 Complete • Retracing The Scale

The Journey From Phone to Silicon

Phone
↓
Chip / SoC
↓
Functional Block
↓
Logic Gates
↓
Transistor
↓
Silicon Atoms

“If all of this computation ultimately reduces to microscopic physical voltages on copper wires and silicon channels...”

HOW CAN ELECTRICITY UNDERSTAND 0 AND 1?

How do continuous, noisy analog voltages get transformed into clean, indestructible digital information? In Section 1, we build the mathematical and electrical bridge.