By Cyril KimbiComputer Science splits into a theory paper and a practical one, and candidates almost always prepare for the second while assuming the first will look after itself. It does not. The theory paper contains the number bases, the architecture and the algorithm questions that decide most grades.
Number bases and data representation
Start here. These questions appear on every paper, they are objectively markable, and they are entirely mechanical once practised — which makes them the best return on revision time in the whole subject.
What must be automatic:
- Conversions between denary, binary, octal and hexadecimal, in both directions.
- Binary arithmetic — addition and subtraction, including overflow.
- Signed representation — sign and magnitude, one’s complement, two’s complement, and why two’s complement is used.
- Character encoding — ASCII and Unicode, and why the difference matters.
- Storage units and how a file size is calculated from resolution or sample rate.
Practise these until you can do them without hesitation. Show your working: method marks are awarded even when the final value is wrong, and a bare answer earns nothing if it is incorrect.
Architecture, in the right words
Architecture questions are terminology questions wearing a technical costume. Understanding the idea is necessary and not sufficient — the scheme wants the named component and the named stage.
Be able to state precisely: the components of the CPU and what each does; the fetch-decode-execute cycle as a sequence of named stages; the purpose of each register; the role of buses and what distinguishes address, data and control buses; the memory hierarchy from registers down to secondary storage, and why it exists; and the distinction between primary and secondary storage.
Test yourself in the direction the exam uses — given a description of what something does, produce its name. Most candidates revise the easier direction and are caught out.
Algorithms: write them, then trace them
Algorithm questions ask you to produce or to trace, and the second is where the reliable marks are because it needs discipline rather than invention.
To trace: draw a column for every variable plus one for output, then execute one line at a time, writing each new value in its own row. Do not skip iterations because you can see the pattern — the marks are in the intermediate rows.
To write: know a small set of structures well rather than many badly — linear search, binary search, one sorting algorithm, finding a maximum or minimum, counting and accumulating, and validating input. Almost every question recombines these.
Keep pseudocode consistent. Pick one convention for assignment, input, output and loops and hold it throughout; examiners accept a range of conventions but not three in one answer.
Databases, networks and systems
These sections are definition-dense and reward a list rather than an essay.
Databases: know what a primary key, a foreign key and a relationship are; what normalisation is for; and why a relational database beats a flat file — redundancy, consistency, integrity.
Networks: topologies with labelled diagrams, LAN against WAN, client-server against peer-to-peer, and the common security threats with a countermeasure named for each.
Systems development: the stages of the life cycle in order, and what happens at each. This is a sequence question and is marked as a sequence.
How the two papers differ in preparation
| Theory paper | Practical paper | |
|---|---|---|
| Rewards | Precision and recall | Fluency in producing code |
| Best practice method | Conversions drilled; definitions tested backwards | Writing programs on paper, then tracing them |
| Cheapest marks | Number bases, trace tables | Declarations, loop structure, trace tables |
| Needs a computer | No | Helpful, not essential |
Neither paper needs regular machine access to prepare for, which matters if computer time is limited. Both are written on paper.
A revision routine
- Twenty conversions a week, timed, until they are automatic. Highest return in the subject.
- A definitions list from the syllabus, tested description-to-term.
- Trace one algorithm a week by hand, fully.
- Past papers timed, marked strictly against the scheme, with a logged list of recurring losses.
- Final month: speed only. Nothing new.
The Computer Science past papers supply the questions.
Frequently asked questions
Do I need a computer to pass Computer Science?
No. Both papers are written on paper. A computer helps you verify programs but is not required to prepare.
What is the difference between Computer Science and ICT?
Computer Science covers how computers work and how to program them; ICT covers using and applying technology.
Which topic should I revise first?
Number bases and data representation. They appear on every paper and are mechanical once drilled.
How much pseudocode do I need to memorise?
None verbatim. Know a small set of structures well enough to write them from memory and recombine them.
Are method marks really awarded?
Yes, frequently — which is why you should always show working and always write the part of an algorithm you are sure of.
