Why does a computer with plenty of storage still slow down when it runs too many applications? Storage capacity, main-memory capacity and processor performance describe different resources. A laptop could have a terabyte of storage, yet operate poorly with too little available RAM for its active tasks. The first skill in computer architecture is learning which component is responsible for which job.
Ministry of Education V3.0 Reference 359 teaches the 2027 Singapore–Cambridge SEC G3 Computing K349 section 1.1 on computer architecture. Students learn how processors, main memory and secondary storage work together; how address and data buses enable memory access; what USB, HDMI and PCI Express interfaces do; how magnetic, optical and solid-state storage differ; and how to calculate data quantities using decimal and binary unit prefixes. Worked comparisons turn technical vocabulary into explanations parents, tutors and students can test.
This is an independent eduKateSG guide, not an official MOE, SEAB or Cambridge publication. Examinable scope and assessment conditions come from the official 2027 K349 syllabus. All invented devices and classroom scenarios below are illustrative, not equipment performance benchmarks. Actual speeds, prices and component behaviour vary by generation and model. We focus on accurate relationships and clear evidence rather than unsupported claims that one technology is always best.
Start here: trace one simple computing task
Suppose a student opens a spreadsheet saved on an SSD. The application code and required data are retrieved from secondary storage into main memory as necessary; the processor executes instructions and operates on data; the screen displays results through the display system. When the student saves a file, it is written to persistent storage. That broad path gives you a way to assign each action to a component.
A useful self-test is to answer four questions without looking down: Which component executes instructions? Which holds active instructions and data for quick access? Which retains files when power is removed? Why does a processor need a way to identify particular memory locations? If any answer is uncertain, revisit the corresponding section before attempting detailed unit calculations.
1. Read the official K349 section 1.1 boundaries precisely
The 2027 G3 Computing syllabus specifies five outcomes. Candidates calculate quantities in bits, bytes, kilobytes, kibibytes, megabytes, mebibytes, gigabytes, gibibytes, terabytes, tebibytes, petabytes and pebibytes. They explain processor, main memory and secondary storage; the roles of data and address buses; USB, HDMI and PCI Express interfaces; and magnetic, optical and solid-state storage characteristics.
This article concentrates on those outcomes. The syllabus does not ask students to memorise the internal transistor layout of a particular processor, list the release dates of connector versions or benchmark a commercial laptop. Good preparation understands function, data movement, units and trade-offs. Hardware changes quickly, so explain general principles and treat specific marketed capacities or speeds as examples rather than exam facts.
2. The processor executes instructions; it is not the storage cabinet
The central processing unit, or CPU, fetches and executes program instructions. Its operations can include arithmetic, comparison, control and coordination of data movement. A processor is essential to running programs, but it does not replace main memory or the long-term storage medium. Calling the CPU ‘where everything is saved’ confuses processing with retention.
The CPU may include small, fast internal storage such as registers and caches. These are useful to understand conceptually but should not be mistaken for the syllabus’s broader main-memory versus secondary-storage distinction. A program can still be slow when it waits for data or input/output even if the processor is highly capable. Always ask which part of a task is limiting the outcome.
3. Main memory is the active working area
Random access memory is a form of main memory commonly used for the instructions and data of running programs. It allows the processor to access working information quickly. Ordinary RAM is volatile: its contents are not preserved merely by switching off the computer. Students should distinguish this property from how much data the memory can hold.
A computer with inadequate RAM may have to move information between main memory and slower secondary storage more often, potentially reducing responsiveness. More RAM can help when memory capacity is the bottleneck, but adding RAM does not guarantee that every task becomes faster. The useful explanation identifies the resource involved and the workload, rather than claiming that one component determines all performance.
4. Secondary storage preserves programs and files
Secondary storage is used for persistent data such as documents, applications and operating-system files. Common media include magnetic hard disks, optical discs and solid-state drives. They differ in physical mechanism, durability, portability, typical capacities, cost and speed. A correct comparison names the dimension being judged, rather than treating ‘better’ as a universal property.
An SSD generally avoids the moving read/write heads of a hard disk and offers quick random access. A hard disk can provide a lower cost per unit of capacity in some contexts. Optical discs can be useful for distribution or archival tasks, but their capacity, write characteristics and vulnerability to physical damage matter. Specific product choices depend on contemporary specifications, so avoid fixed claims that every SSD is faster than every HDD in all operations.
5. Understand how an address differs from the data stored there
A memory address identifies a location the processor wants to access. The address bus carries information specifying the relevant location. The data bus carries the actual data exchanged during a read or write. The difference is analogous to specifying a shelf position and moving a book: the position describes where; the book is what is transferred.
The analogy is imperfect—computer memory uses electronic mechanisms rather than shelves—but it preserves the distinction. If a question asks which bus identifies the desired memory location, answer address bus. If it asks which carries the value read from that location, answer data bus. Mixing these names is a frequent conceptual error because both buses participate in the same operation.
6. Follow a memory read step by step
In a simplified model, the processor identifies a memory location to read. The location address is presented via the address bus. Control signals indicate the required operation. The memory subsystem makes the stored data available, and the value is transferred to the processor via the data bus. The control bus is not the focus of this syllabus outcome, though real systems require coordination.
A student should not overinterpret this simplified model as an exact timing diagram for all modern CPUs. Cache hierarchies and memory controllers complicate practical implementation. For the examination, explain the functional roles accurately. Then test the answer: if the processor asks for a different location but the stored value happens to be identical, the address would change while the resulting data could remain the same.
7. A memory write moves a value to a specified location
During a write, the processor or system provides the address of the target location and the data that should be stored there, with appropriate control. The address tells the memory where to place the new value; the data bus conveys what that value is. This is different from reading, where the required value moves back to the requester.
When explaining a read or write, do not confuse the direction of information flow with the purpose of each bus. The address bus is about selecting a location, not announcing whether the number being transferred is large or small. The data bus can carry values that represent numbers, characters or parts of instructions; the bit pattern’s meaning depends on the wider program.
8. Interfaces connect systems but perform distinct roles
USB provides a widely used interface for peripherals such as storage devices, keyboards and cameras, although its capabilities vary by connector and supported protocol. HDMI is an interface commonly used for digital video and audio output to displays. PCI Express is a high-speed internal expansion and interconnect standard used for devices such as graphics cards and some SSDs.
Avoid using connector shape as the only way to identify capability. USB-C describes a physical connector, while the particular device and cable determine which features and data rates are supported. HDMI versions and implementations also differ. PCI Express uses lanes and generations that affect available throughput. In a syllabus answer, explain typical application, connector and comparative speed rather than memorising unverified retail marketing claims.
9. A connector is not the same thing as a complete computer system
A display cable does not execute drawing instructions; it transports signals to a suitable display. A storage interface does not decide the contents of an essay; it supports data transfer. An input/output interface serves an exchange between compatible devices, constrained by the hardware, protocol and installation.
For example, a student cannot assume that a USB socket will always provide the same transfer rate across all computers or cables. Similarly, a PCI Express card needs compatible system resources and drivers. Good architecture literacy asks about purpose, compatibility, bandwidth and the required task. These questions are more durable than remembering an isolated speed advertised for one product.
10. Distinguish capacity from speed
A storage device may hold more data but transfer it more slowly. Main memory can have a large capacity yet still be used inefficiently by a particular workload. Processor clock rates are not sufficient on their own to compare overall computing performance. Capacity answers ‘How much?’; throughput answers ‘How much per unit time?’ and latency asks ‘How long until the response begins?’
The three can influence one another but are not interchangeable. A family buying a computer for word processing may need modest capacities and reliable responsiveness. Someone editing large video files has different demands. In an exam response, identify the relevant measure before comparing technologies. Avoid unsupported claims that increasing capacity always increases speed.
11. Data quantities: the exact difference between bits and bytes
A bit has two possible values, and eight bits make one byte. A file of 2,000 bytes contains 16,000 bits. Be careful with lower-case b and upper-case B in unit notation: Mb commonly means megabits, while MB commonly means megabytes. That distinction creates an eightfold difference even before decimal and binary prefixes are considered.
For instance, a hypothetical link transferring 80 megabits per second has a theoretical data-rate equivalent of 10 megabytes per second if decimal units are used and no overhead is considered. Actual file transfer is affected by protocol overhead, system performance and network conditions. The educational calculation is only a unit conversion, not a promise of the transfer speed a customer will experience.
12. Decimal prefixes use powers of 1,000
One kilobyte is 1,000 bytes; one megabyte is 1,000,000 bytes; one gigabyte is 1,000,000,000 bytes. The pattern continues through terabytes and petabytes. The decimal prefixes fit SI usage and are common in manufacturers’ storage specifications. Always read the unit stated in the question before choosing a multiplier.
For a simple calculation, 2.5 MB equals 2,500,000 bytes, or 20,000,000 bits. If you need to express it in kilobytes, divide by 1,000 to obtain 2,500 kB. Do not switch casually from base-1,000 prefixes to base-1,024 prefixes in the middle of the same calculation. Consistency is part of correctness.
13. Binary prefixes use powers of 1,024
One kibibyte, or KiB, equals 1,024 bytes. One mebibyte, or MiB, equals 1,024 KiB, which is 1,048,576 bytes. One gibibyte equals 1,024 MiB, and so on. The syllabus explicitly names the corresponding binary prefixes through pebibytes, so students should practise their distinction from decimal kilobytes, megabytes and larger quantities.
Take an amount of 1,048,576 bytes. That is precisely 1 MiB but is 1.048576 MB when expressed using decimal megabytes. Both descriptions refer to the same quantity, not two different files. A likely exam error is calling 1,024 bytes ‘one kilobyte’ without checking the convention. Use the explicit labels when the question provides them.
14. Convert between units with a written chain
If a file is 4 MiB, multiply by 1,024 twice to obtain 4,194,304 bytes; multiply by eight to obtain 33,554,432 bits. For a decimal quantity of 4 MB, the corresponding values are 4,000,000 bytes and 32,000,000 bits. The numbers are different because the prefixes differ.
A reliable solution writes every unit alongside the numerical operation. Rather than jumping directly from an unfamiliar unit to bits, show a dimensional chain. When dividing to convert bytes into a larger unit, ask whether the output should become numerically smaller. If a calculation turns 1,000 bytes into a million megabytes, directionality already reveals the mistake.
15. Compare magnetic and solid-state media with criteria
Magnetic hard-disk drives store information using magnetised surfaces and mechanical components. Solid-state drives use electronic flash memory. Depending on the models compared, SSDs often have faster access and greater resistance to movement-related damage while HDDs may offer lower cost per capacity. The appropriate answer depends on workload, budget and portability needs.
No storage medium is immune to failure. SSDs have endurance and controller considerations, while HDDs face wear of moving parts and sensitivity to shocks. Backups remain necessary regardless of device type. A good storage recommendation names expected usage and the evidence that a particular trade-off matters, not the assumption that ‘newest’ means universally safest.
16. Optical media and portability require context
Optical discs such as CDs, DVDs and Blu-ray discs are read using a laser. Their typical capacities and read/write characteristics differ by format. Portability depends on the availability of compatible drives as well as the physical disc itself. A disc may be cheap to distribute yet inconvenient when current computers lack optical drives.
Students should distinguish a storage technology’s properties from an intended use. A format that suits fixed archival distribution may perform poorly for frequently updated working files. Evaluation might involve storage capacity, speed, robustness, ease of transport and replacement cost. Every comparison needs an actual scenario to avoid meaningless rankings.
17. Performance bottlenecks make the component roles visible
Imagine a computer editing a large collection of photos. The CPU executes instructions, RAM holds active work, storage supplies files and retains saved results, while input and display interfaces connect equipment. If the job is slow, the cause could be processing limits, insufficient memory, slow I/O or software behaviour. You cannot identify the cause merely from the file’s size.
A useful classroom investigation would compare observed behaviour under controlled tasks on school-approved equipment. It should not require opening electrical hardware or manipulating unsafe components. The learner can report what changed, what stayed constant and what remains uncertain. That method turns architecture vocabulary into diagnosis.
18. The twelve-stage Task–Component–Unit checking protocol
This is a teaching routine, not an official grading rubric. It helps students turn a story about a computer into the relevant architectural concept, then check calculations and trade-offs without being distracted by unfamiliar product names.
Stage 1 — Name the task
Identify whether the device is executing, retaining, displaying or transferring information. Keep the verb in the question visible.
Stage 2 — Identify the processor’s role
Determine which part performs the requested computation or coordinates execution. Avoid assigning persistent file storage to the processor.
Stage 3 — Identify main memory
Ask what information must remain immediately accessible while programs run. Distinguish fast active working data from persistent files.
Stage 4 — Identify secondary storage
Find where files and installed programs remain when the system is off. Note the actual retention requirement.
Stage 5 — Trace the address bus
If a memory location is being accessed, state that the address identifies the requested location.
Stage 6 — Trace the data bus
State which value is transferred and the direction appropriate to reading or writing.
Stage 7 — Recognise the interface
Distinguish USB peripheral links, HDMI display output and PCI Express expansion roles in the stated scenario.
Stage 8 — Name the physical medium
Identify whether the secondary storage described is magnetic, optical or solid-state.
Stage 9 — Select a comparison criterion
Choose speed, capacity, portability, durability or cost and explain why that criterion matters to the task.
Stage 10 — Write the unit convention
Identify bit versus byte and decimal versus binary prefixes before calculating.
Stage 11 — Check the magnitude
Verify the arithmetic through an inverse conversion or a quick reasonableness test, including whether a unit conversion increases or decreases the number.
Stage 12 — State a justified conclusion
Connect the component or technology to the problem without claiming universal performance or certainty unsupported by evidence.
19. Twenty original computer-architecture mini-cases
Each mini-case is fictional and focuses on the reasoning, not an official examination question. For the calculation examples, use the explicit unit convention and check the result before reading the explanation. For the component questions, identify which statement describes a function and which merely repeats a marketing label.
Case 1 — The student opens a saved project
A file resides on persistent storage until the computer needs its contents for active work. Relevant data can be loaded into main memory, and the processor executes instructions. Saying ‘the CPU stores the essay permanently’ confuses execution with retention.
Case 2 — The computer loses power during editing
Unsaved changes held only in volatile working memory can be lost, although software may have saved recovery data to persistent storage. The educational distinction concerns volatility and persistence, not a guarantee about one application.
Case 3 — An SSD has a larger capacity but feels slow
Capacity is not throughput. Performance can depend on the device, interface, remaining workload and software. Identify the operation being timed before claiming that a large SSD must be faster.
Case 4 — A graphics card is installed in an expansion slot
PCI Express is a suitable high-speed internal interconnect for many graphics cards. Explain its role as a bus/interface rather than treating it as a file format or video display itself.
Case 5 — A classroom display shows a blank screen
HDMI can carry digital audio and video to a compatible monitor. A blank screen could also be caused by source selection, power, cable or settings. Do not attribute every display failure to the CPU.
Case 6 — A USB device copies files slowly
USB transfer performance can depend on the supported standard, device, cable, source storage and workload. The connector shape alone does not prove the maximum speed. Explain which factor is being investigated.
Case 7 — A processor reads memory location 200
The address bus communicates the selected location, while the data bus carries the returned value. The value held at location 200 is not necessarily the number 200.
Case 8 — A processor writes a new value to memory
A write involves an address to identify where, and data to identify what. If a question changes only the value to be written, that does not mean the target address must also change.
Case 9 — A computer needs more active applications
Additional RAM can help if memory capacity is genuinely constraining the workload. But it cannot automatically accelerate a program limited by CPU computation or an external network service.
Case 10 — A user chooses an external hard disk
A magnetic disk can provide substantial persistent capacity, while performance and shock sensitivity depend on model and handling. The correct answer compares requirements rather than assuming any external device is universally safe.
Case 11 — An optical archive must be reopened next year
Optical media may be appropriate in some settings, but future drive availability, disc condition and storage procedures matter. Retention is a system problem, not just a property of the disc.
Case 12 — A memory stick is removed without saving
A portable USB flash drive is solid-state secondary storage. Whether recent edits are present depends on whether the application wrote them successfully, not on the fact that the storage is called ‘memory’.
Case 13 — Convert 2.5 MB to bytes
Under decimal prefixes, 2.5 MB = 2,500,000 bytes. Multiply by eight for 20,000,000 bits. This is a unit relationship rather than a device-speed estimate.
Case 14 — Convert 4 MiB to bytes
Four mebibytes equal 4 × 1,024 × 1,024 = 4,194,304 bytes. An answer of 4,000,000 bytes would correspond to 4 MB, not 4 MiB.
Case 15 — Convert 10 kB to bits
Ten decimal kilobytes contain 10,000 bytes, hence 80,000 bits. Write the two multipliers explicitly: 1,000 bytes per kB and eight bits per byte.
Case 16 — Convert 3 KiB to bits
Three kibibytes equal 3 × 1,024 = 3,072 bytes, hence 24,576 bits. The binary prefix determines the first factor.
Case 17 — Compare 1 GB and 1 GiB
One gigabyte is 1,000,000,000 bytes. One gibibyte is 1,073,741,824 bytes. The latter is larger by 73,741,824 bytes, even though both may be casually described as ‘one gigabyte’ in informal speech.
Case 18 — The student confuses Mbps and MB/s
Eighty megabits per second is 10 megabytes per second as a theoretical unit equivalence using decimal notation. Actual application throughput may be lower. Capitalisation changes the unit by a factor of eight.
Case 19 — A school compares laptop designs
One device has faster secondary storage; another has more RAM. The better choice depends on whether students mostly open large files, run memory-intensive software or carry the device daily. Describe a relevant workload.
Case 20 — A repair question names a broken address bus
A failure affecting correct location selection would impair memory access. By contrast, a data-transfer problem concerns conveying values. The exam answer should explain the conceptual distinction, not invent an exact hardware diagnosis.
20. Practice set: diagnose, calculate and explain
Question one: a fictional device has 8 GiB of RAM and 256 GB of SSD space. Which quantity describes main memory and which describes secondary storage? Question two: if the processor requests a word from one memory address, which bus identifies the location? Question three: explain why HDMI and PCI Express should not be treated as equivalent ports. Question four: convert 5 MiB into bytes and bits, showing working.
Question five: compare magnetic and solid-state media for a portable computer used frequently while travelling, naming at least three criteria and one uncertainty. Question six: explain what a terabyte and tebibyte represent and why they differ. Answers: 8 GiB is RAM, 256 GB is SSD space; the address bus identifies location; HDMI typically serves display output while PCI Express is an internal expansion interconnect; 5 MiB = 5,242,880 bytes = 41,943,040 bits.
21. A focused seven-day learning sequence
Day one, draw a simple component map with processor, RAM, storage, input and output. Day two, explain a read and write using the two buses. Day three, compare USB, HDMI and PCI Express applications. Day four, contrast magnetic, optical and solid-state media using scenarios. Day five, practise decimal prefixes. Day six, practise binary prefixes. Day seven, mix the questions without chapter labels.
Tutors can use a miniature oral interview: ‘A school notebook has enough disk space but fails to run several programs smoothly. What could matter, and what could you not conclude yet?’ A student who names a plausible limitation while acknowledging alternatives is thinking more rigorously than one who automatically recommends buying a faster processor. Parents can support short retrieval practice, but not unsupervised hardware modification.
22. Frequently asked questions
These questions summarise the distinctions most likely to be confused. The correct answer is often the one that states the resource and task precisely rather than claiming one component is always better.
Is RAM the same thing as storage?
Both hold information, but they serve different roles. Main memory supports active processing and is typically volatile; secondary storage retains programs and files persistently. The advertised capacities describe different resources.
Does a bigger SSD make the CPU faster?
No. Faster access to files can reduce some delays, but the processor’s inherent computational capabilities do not automatically change when a storage device is replaced.
What does an address bus carry?
It carries information identifying the memory location to be accessed. It is distinct from the data bus, which carries the value read or written.
What does the data bus carry?
It carries data exchanged among system components during operations such as memory reads and writes. The address bus selects a location; the data bus moves values.
Is USB-C always faster than older USB connectors?
No. USB-C is a connector form; the supported protocol, cable and connected device determine capabilities. Some devices with the same connector can have different transfer rates.
Is HDMI a storage interface?
Its typical purpose is transmitting digital video and audio to displays and related devices, not persisting file data.
What is PCI Express typically used for?
It is a high-speed internal interconnect widely used by expansion devices, including some graphics cards and solid-state storage. The available lane configuration and generation affect performance.
Why are MB and MiB different?
MB uses decimal powers of 1,000 and MiB uses binary powers of 1,024. One MiB contains 1,048,576 bytes while one MB contains 1,000,000.
Does an SSD retain information when switched off?
Yes, under normal operating conditions, solid-state drives provide non-volatile secondary storage. Nevertheless, failures are possible and important files should have backups.
Which storage type is the most durable?
There is no universal answer. Physical handling, environment, device quality, write usage, compatibility and maintenance differ. Evaluate a specified scenario rather than treating the medium name as a guarantee.
Official syllabus and connected eduKateSG reading
This guide follows the official 2027 SEC G3 Computing K349 syllabus, section 1.1 (Computer Architecture); the SEAB 2027 G3 subject directory gives the current entry point. Continue with Reference 358 — Data Representation, Reference 306 — Networks, the MOE V3.0 learning library and the immutable Clementi teaching reference.
Reviewed 29 October 2026, Singapore. All numerical scenarios are original classroom illustrations; advertised specifications and costs of real devices are subject to change. The enduring habit is to identify the component, trace the information flow, name the quantity and check the units before choosing a conclusion.
MOE V3.0 reference navigation
Previous: Reference 358 — Computing Data Representation · MOE V3.0 Library.