Why does your 100Mbps internet only download at 12MB/s? Why does a newly purchased 1TB hard drive show only 931GB? Why can a 32-bit system only use 3.2GB of RAM? What happens when you open an application — how do CPU, memory, and disk collaborate? This is the first part of the Computer Fundamentals Deep Dive Series, where we'll start from the most basic data units (Bit/Byte), dive deep into CPU working principles, explore Intel vs AMD architectural differences, understand server-grade processors, and learn how to select the right CPU for your needs. Through extensive real-world analogies and practical cases, you'll truly understand how your computer's "brain" operates.

📚 Computer Fundamentals Deep Dive Series (5 Parts): 1. → CPU & Computing Core (Data units, processor architecture, Intel vs AMD) ← You are here 2. Memory & High-Speed Cache (DDR evolution, dual-channel, L1/L2/L3 cache) 3. Storage Systems Complete Analysis (HDD vs SSD, interfaces, RAID) 4. Motherboard, Graphics & Expansion (PCIe, USB, GPU, BIOS) 5. Network, Power & Practical Troubleshooting (NICs, PSU, cooling, diagnostics)

Opening: Three Real-World Scenarios

Scenario 1: The Bandwidth Mystery

You subscribed to 100M broadband
Download speeds show 12MB/s
Question: What's the relationship between 100M and 12MB/s?

Scenario 2: Missing Storage Space

Purchased a Western Digital 1TB HDD
Windows shows 931GB available
Truth: The 69GB loss is due to 1000 vs 1024 calculation differences

Scenario 3: Game FPS Disparity

Your friend's i5-13600K runs CS:GO at 450 FPS
Your R9 7950X (more cores!) only gets 420 FPS
Reason: Games favor single-core performance, not multi-core count

Part 1: Data Units - The Metric System of Computing

Bit & Byte: The Smallest Units

What is a Bit?

Definition: The smallest unit of data in computing, having only 0 or 1 as states.

Real-world analogy: A bit is like a light switch

0 = Off (no current)
1 = On (current flowing)

8 switches (8 bits) can representdifferent combinations.

Practical applications:

Network speed: 100 Mbps = 100 megabits per second
Download speed: 12.5 MB/s = 12.5 megabytes per second
Conversion: Common misconception: > ❌ "My ISP advertised 100M internet, why am I only getting 12MB/s? Am I being scammed?" > > ✅ ISPs advertise in Mbps (megabits), your browser shows MB/s (megabytes). There's an 8x difference!

Pro tip: To convert advertised speed to download speed, divide by 8.

What is a Byte?

Definition: 1 Byte = 8 Bits, the fundamental unit for measuring storage capacity.

Why 8 bits?

Historical reason:

Early computers used 8 bits to represent a character (ASCII) -combinations = enough for all English letters, digits, punctuation
Example: Letter A has ASCII code 65 = binary 01000001 (8 bits)

Modern encoding:

ASCII: 1 character = 1 byte (English letters, numbers)
UTF-8: 1 Chinese character = 3 bytes
Unicode: 1 emoji = 4 bytes

Example:

String "Hello World" takes how many bytes?

H e l l o   W o r l d   →  11 bytes (ASCII)
Space       →  1 byte
(CJK chars) →  6 bytes (UTF-8, 3 bytes per character)
Total       → 18 bytes

Unit Conversion: The Secret of 1024

Standard Conversion (Computer Internal)

Unit	Conversion	Exact Value (Bytes)	Magnitude
1 KB	Byte	1,024 Byte	Thousand
1 MB	Byte	1,048,576 Byte	Million
1 GB	Byte	1,073,741,824 Byte	Billion
1 TB	Byte	1,099,511,627,776 Byte	Trillion
1 PB	Byte	-	Quadrillion
1 EB	Byte	-	Quintillion

Why 1024 instead of 1000?

Because computers use binary:

Closest power of 2 to 1000 is
Makes binary arithmetic efficient (bit-shifting)

Analogy:

Clocks use base-60 (60 seconds = 1 minute)
Computers use base-1024 (1024 bytes = 1KB)

Manufacturer Conversion (Disk Labeling)

Unit	Manufacturer	Exact Bytes
1 KB	1,000 Byte	1,000
1 MB	1,000 KB	1,000,000
1 GB	1,000 MB	1,000,000,000
1 TB	1,000 GB	1,000,000,000,000

Why do manufacturers use 1000?

Simpler calculations: 1000 is easier to compute mentally
Legally permitted: International System of Units (SI) defines kilo = 1000
Marketing strategy: Same drive appears "larger" in base-1000

Real-World Case: Where Did My Storage Go?

Scenario: 1TB Drive Shows 931GB

Purchase: Western Digital 1TB HDD, box says 1TB

After installation: Windows Explorer shows 931GB available

User confusion: Where did the missing 69GB go? Is the drive defective?

Calculation breakdown:

Manufacturer calculation (base-1000):

System calculation (base-1024):

Loss percentage:

Quick estimation formula: > Actual capacity ≈ Advertised capacity × 0.931

Reference table:

Advertised	Actual (GB)	Loss
128 GB	119 GB	9 GB (7.0%)
256 GB	238 GB	18 GB (7.0%)
500 GB	466 GB	34 GB (6.8%)
1 TB	931 GB	69 GB (6.9%)
2 TB	1,863 GB	137 GB (6.9%)
4 TB	3,726 GB	274 GB (6.9%)
8 TB	7,452 GB	548 GB (6.9%)

Conclusion: This isn't a defect — it's a conversion standard difference!

Q&A: Common Unit Conversion Questions

Q1: Why doesn't RAM "shrink" like hard drives do?

A: Because RAM manufacturers use base-1024, while hard drive manufacturers use base-1000!

RAM: 8GB stick =bytes ✅ No shrinkage
HDD: 1TB drive =bytes ⚠️ Shrinks by 6.9%

Q2: With 100M broadband, what download speed should I expect?

A: Theoretical maximum =Actual speeds typically 10-11 MB/s due to:

TCP/IP protocol overhead (~5-10%)
Network congestion and packet loss
Server-side rate limiting

Real example:

Advertised Speed	Theoretical Download	Actual Download (Typical)
100 Mbps	12.5 MB/s	10-11 MB/s
200 Mbps	25 MB/s	20-23 MB/s
500 Mbps	62.5 MB/s	55-60 MB/s
1 Gbps	125 MB/s	110-120 MB/s

Q3: Why does formatting reduce capacity further?

A: File systems need to store metadata:

File Allocation Table (FAT): Tracks file locations
Directory structure: Folder tree hierarchy
Journaling (NTFS/ext4): Ensures data consistency

Typically consumes 1-2% of capacity.

Example:

1TB drive actual: 931GB
After NTFS formatting: ~915GB available
Filesystem metadata: ~16GB

Q4: What's the difference between GB and GiB?

A: GiB (Gibibyte) = precise binary unit!

Symbol	Name	Calculation	Value
GB	Gigabyte	bytes (sometimes)	Ambiguous
GiB	Gibibyte	bytes	1,073,741,824 bytes

Recommendation: Use GiB for precision, accept that GB is ambiguous in practice.

Memory Cheat Sheet

Computer uses 1024, manufacturers use 1000;

Broadband in bits, download in bytes, divide by 8 for real speed;

TB to GB multiply 0.931, hard drive shrinkage is normal!

Part 2: CPU - The Computer's Brain

CPU's Core Responsibilities

What Does CPU Do?

Simple definition: CPU = Central Processing Unit, responsible for executing instructions and processing data.

Analogy: CPU is like a company's CEO

Receives tasks: Reads instructions from memory
Makes decisions: Performs calculations (arithmetic, logic)
Delegates work: Coordinates disk, GPU, network card, etc.

Data Flow Path

User Input (keyboard/mouse)
   ↓
Input Device → Memory (temporary storage)
   ↓
CPU reads from memory → Executes computation
   ↓
CPU writes results → Memory
   ↓
Memory → Output Device (monitor/printer)

Example: Typing =SUM(A1:A10) in Excel

Input: Keyboard types formula → Memory
Processing: CPU reads A1-A10 values from memory → Calculates sum
Output: CPU writes result to memory → Monitor displays

Why Do We Need Memory?

Key question: Why not read directly from disk?

Answer: Massive speed difference!

Storage Device	Access Latency	Relative Speed	Analogy
CPU Register	~0.1ns	1x	Your pocket (instant access)
L1 Cache	~1ns	10x	Notebook on desk
L2 Cache	~4ns	40x	Bookshelf in room
L3 Cache	~15ns	150x	Office filing cabinet
RAM (DDR4)	~100ns	1,000x	Adjacent room storage
SSD (NVMe)	~100μ s	1,000,000x	Company warehouse
HDD (Mechanical)	~10ms	100,000,000x	City library across town

Visualization: If CPU thinking takes 1 second:

Fetching from memory = 17 minutes (tolerable)
Reading from SSD = 11.5 days (too slow)
Reading from HDD = 3.2 years (completely unusable)

So memory is an essential intermediary between CPU and disk!

CPU Brand Battle: Intel vs AMD

Intel: The Established Leader

Product lines:

Consumer: Core i3 / i5 / i7 / i9
Server: Xeon (enterprise-grade)

Strengths:

✅ Strong single-core performance: Better gaming
✅ Mature ecosystem: Excellent software optimization
✅ High stability: Enterprise first choice
❌ Higher prices: Lower price-to-performance ratio
❌ Fewer cores: Lags behind AMD in multi-core

Target users:

Gamers (single-thread sensitive games)
Enterprise users (stability priority)
Users with generous budgets

AMD: The Value Champion

Product lines:

Consumer: Ryzen 3 / 5 / 7 / 9
Server: EPYC (server-grade, high core count)

Strengths:

✅ Strong multi-core performance: Excels in multitasking, video rendering
✅ Better value: More cores at same price point
✅ Excellent efficiency: Advanced 7nm/5nm process
❌ Slightly weaker single-core: Some games have lower FPS
❌ Driver optimization: Some professional software less optimized

Target users:

Content creators (video editing, 3D rendering, compilation)
Multitasking users (running many applications simultaneously)
Budget-conscious performance seekers

Performance Comparison (2024 Mainstream Models)

CPU Model	Cores/Threads	Single-Core	Multi-Core	Price	Best For
Intel i5-13600K	14C/20T	★★★★★	★★★★☆	$280	Gaming + Light Creation
AMD R7 7700X	8C/16T	★★★★☆	★★★★★	$260	All-rounder
Intel i7-13700K	16C/24T	★★★★★	★★★★★	$400	High-end Gaming + Creation
AMD R9 7900X	12C/24T	★★★★☆	★★★★★	$380	Professional Creation

Selection guide:

Use Case	Recommended CPU	Reason
Gaming-focused	Intel i5/i7	High single-core frequency
Video editing	AMD R7/R9	Multi-core advantage
3D rendering	AMD R9/Threadripper	Maximum core count, parallel processing
Office work	Intel i3 / AMD R3	Sufficient and economical
Server	AMD EPYC	High core density, better value

32-bit vs 64-bit: Memory Ceiling

Basic Comparison

Feature	32-bit CPU	64-bit CPU
Address bus width	32 bits	64 bits
Addressable memory (theory)	GB	EB
Actually usable memory	~3.2 GB	Virtually unlimited (motherboard limited)
Mainstream OS	Windows 7 32-bit (obsolete)	Windows 10/11 64-bit
Software compatibility	32-bit only	Both 32-bit & 64-bit ✅

Why Only 3.2GB Usable with 4GB Theoretical?

Answer: Memory-Mapped I/O (MMIO) occupies part of the address space!

Detailed explanation:

A 32-bit address bus can accessaddresses (1 byte each), theoretically 4GB.

But this 4GB address space isn't all allocated to RAM. It must also map:

Graphics VRAM: 512MB - 2GB
BIOS chip: Several MB
PCI devices: Network card, sound card registers

Address space allocation diagram:

32-bit system's 4GB address space:

0x00000000 ─────┐
               │
               │  System RAM
               │  ~3.0-3.5GB usable
               │
               │
0xC0000000 ─────┤
               │
               │  MMIO Region
               │  (Graphics VRAM mapping)
               │  (PCI device registers)
               │  ~0.5-1.0GB
               │
0xFFFFFFFF ─────┘

Real case study:

Installing 4GB RAM on 32-bit Windows 7:

System Properties shows: "Installed memory: 4.00GB"
But displays: "Usable memory: 3.25GB"
Reason: 0.75GB reserved for graphics and device address mapping

Solution: Upgrade to 64-bit system!

With 64-bit, address space is= 16 EB (exabytes), memory ceiling effectively removed.

PAE (Physical Address Extension)

What is PAE?

PAE is a feature allowing 32-bit systems to access more than 4GB RAM (up to 64GB).

How it works:

Extends address bus from 32 bits to 36 bits -GB addressable
Used in server editions (Windows Server 2003/2008)

Limitations:

❌ Single process still limited to 2-3GB
❌ Poor software support
❌ Compatibility issues

Conclusion: PAE is a workaround, not a real solution. Just use 64-bit!

Server CPUs: Enterprise-Grade Performance

Intel Xeon: The Enterprise Standard

Core features:

ECC Memory Support:
- ECC = Error-Correcting Code
- Automatically detects and corrects single-bit errors
- Critical for: Financial transactions, medical data, scientific computing
Multi-socket capability:
- Supports 2-8 CPUs working together
- Example: Dual Xeon Gold 6248R (48 cores × 2 = 96 cores)
- Applications: Large databases, virtualization platforms
More PCIe lanes:
- Consumer CPU: 16-20 PCIe lanes
- Xeon: 48-64 PCIe lanes
- Benefit: Simultaneously use multiple GPUs, NICs, NVMe drives

Pricing:10,000+

Analogy:

Desktop CPU = Sedan (sufficient for daily use)
Xeon = Bus (carries more passengers, but expensive and fuel-hungry)

AMD EPYC: The Value Challenger

Key advantages:

Higher core density:
- AMD EPYC 7763: 64 cores / 128 threads
- Similar-priced Xeon: 48 cores / 96 threads
- Improvement: +33% more cores
Greater memory bandwidth:
- 8-channel DDR4 (Xeon typically 6-channel)
- Better for memory-intensive workloads (databases, VMs)
Cost advantage:
- Similar performance at 20-30% lower price

Disadvantages:

Ecosystem less mature than Intel
Some enterprise software (Oracle, SAP) less optimized

Huawei Kunpeng: Domestic Alternative

Characteristics:

Based on ARM architecture (different from x86)
Excellent power efficiency (lower power consumption)
Localization solution (information security)

Application scenarios:

Government cloud (data sovereignty)
5G base stations (edge computing)
High security-requirement enterprises

Disadvantages:

Software ecosystem less mature than x86
Performance gap still exists (catching up)

Q&A: Common CPU Questions

Q1: Are more cores always better?

A: Not necessarily! Depends on workload!

Real-world benchmark (Gaming vs Rendering):

Task Type	i5-13600K (14 cores)	R9 7950X (16 cores)	Winner?
CS:GO Gaming	450 FPS	420 FPS	i5 ✅ (stronger single-core)
Premiere Pro Export	5.2 min	4.1 min	R9 ✅ (more cores)
Blender Rendering	8.3 min	6.1 min	R9 ✅ (parallel rendering)
Word/Excel Office	Smooth	Smooth	Tie (both sufficient)

Conclusion:

Gaming/single-threaded → Prioritize single-core performance
Rendering/multitasking → More cores is better
Office work → 4 cores sufficient

Q2: Why do server CPUs have such low frequencies?

A: Servers prioritize stability and sustained high load, not burst performance.

Comparison:

Desktop i9-13900K: Max turbo 5.8GHz (short bursts, high power)
Server Xeon Gold: All-core 3.2GHz (24/7 stable operation)

Analogy:

Desktop CPU = Sprinter (fast bursts, can't sustain)
Server CPU = Marathon runner (steady, sustainable pace)

Why lower frequency is better for servers:

Lower power consumption: 3.2GHz consumes less than 5.8GHz
Better thermal management: Easier to cool in dense server racks
Reliability: Lower voltage = less chance of electromigration failure
Cost-effectiveness: More cores at lower frequency cheaper than fewer cores at high frequency

Q3: Can I upgrade my CPU?

A: Depends on socket type!

Platform	CPU Socket	Upgrade Path	Notes
Intel 12/13 gen	LGA 1700	i3 → i5/i7/i9 (same gen)	Cross-gen not supported
Intel 14 gen	LGA 1700	Compatible with 12/13 gen ✅	Requires BIOS update
AMD Ryzen 5000	AM4	R3 → R5/R7/R9 (even cross-gen)	Great compatibility ✅
AMD Ryzen 7000	AM5	Promised support until 2027	Future-proof investment
Laptop	BGA soldered	Cannot upgrade ❌	Soldered directly to motherboard

Recommendations:

Desktop: Buy the best CPU you can afford initially, or choose AMD AM5 for upgradeability
Laptop: CPU cannot be upgraded — consider long-term needs when purchasing

Memory Cheat Sheet

CPU is brain, RAM is desk, disk is warehouse - can't use directly;

Intel strong single-core great gaming, AMD many cores fast rendering;

32-bit max 4GB minus MMIO, 64-bit address space practically unlimited;

Server multi-socket ECC protects data, desktop high frequency boosts FPS!

What's Next?

In Computer Fundamentals (2): Memory & High-Speed Cache, we'll explore:

Memory working principles: Why do we need RAM between CPU and disk?
DDR generation evolution: From DDR2 to DDR5, how much faster?
Dual-channel real tests: Is 2×8GB really faster than 1×16GB? (with benchmarks)
Three-level cache deep dive: Roles of L1/L2/L3 and hit rates
Memory troubleshooting: Black screen on boot? Is blue screen a memory issue?
Memory optimization: Timings, overclocking, OP reservation

Thought question: If CPU already has L1/L2/L3 cache, why do we still need RAM? Answer in next part!

Series Navigation

Opening: Three Real-World Scenarios

Part 1: Data Units - The Metric System of Computing

Bit & Byte: The Smallest Units

What is a Bit?

What is a Byte?

Unit Conversion: The Secret of 1024

Standard Conversion (Computer Internal)

Manufacturer Conversion (Disk Labeling)

Real-World Case: Where Did My Storage Go?

Scenario: 1TB Drive Shows 931GB

Loss percentage:

Q&A: Common Unit Conversion Questions

Memory Cheat Sheet

Part 2: CPU - The Computer's Brain

CPU's Core Responsibilities

What Does CPU Do?

Data Flow Path

Why Do We Need Memory?

CPU Brand Battle: Intel vs AMD

Intel: The Established Leader

AMD: The Value Champion

Performance Comparison (2024 Mainstream Models)

32-bit vs 64-bit: Memory Ceiling

Basic Comparison

Why Only 3.2GB Usable with 4GB Theoretical?

PAE (Physical Address Extension)

Server CPUs: Enterprise-Grade Performance

Intel Xeon: The Enterprise Standard

AMD EPYC: The Value Challenger

Huawei Kunpeng: Domestic Alternative

Q&A: Common CPU Questions

Memory Cheat Sheet

What's Next?

Further Reading