Cambridge AS Computer Science: Deep Dive into Past Papers | 剑桥 AS 计算机科学：历年真题深度解析

📚 Cambridge AS Computer Science: Deep Dive into Past Papers | 剑桥 AS 计算机科学：历年真题深度解析

Past papers are the most powerful revision tool for Cambridge AS Computer Science (9618). By systematically working through recent exam sessions, you uncover recurring question patterns, key command words, and the precise depth of explanation expected by examiners. This guide dissects the major topics from Paper 1 (Theory) and Paper 2 (Programming) and shows how to turn common errors into full-mark answers.

历年真题是剑桥 AS 计算机科学（9618）最有效的复习工具。通过系统梳理近年的考卷，你会发现重复出现的题型、关键的指令词以及考官期望的解释深度。本文深度解析 Paper 1（理论）和 Paper 2（编程）的各大主题，并展示如何将常见错误转化为满分答案。

1. Understanding the Exam Format & Mark Allocation | 理解考试格式与分值分配

Paper 1 Theory totals 75 marks with a 1 hour 30 minute duration, covering all 12 AS theory topics. Questions are split between short-answer (1–3 marks) and longer structured responses (4–8 marks). Paper 2 Programming Fundamental Problem-solving and Programming is examined through a 2-hour paper worth 75 marks that demands both pseudocode trace-execution and code-writing under time pressure.

Paper 1 理论考试满分为 75 分，时长 1 小时 30 分钟，覆盖全部 12 个 AS 理论主题。题目分为简短答案（1-3 分）和较长的结构化问答（4-8 分）。Paper 2 编程基础问题求解与编程考试为 2 小时的试卷，同样 75 分，要求在时间压力下进行伪代码追踪执行和代码编写。

Examiners repeatedly test the same cognitive skills: knowledge, comprehension, application, and analysis. Prediction is risky, but certain areas – such as CPU architecture, binary arithmetic, network protocols and algorithm tracing – appear every session. You should aim to complete full papers under timed conditions at least four times before the exam.

考官反复测试相同的认知技能：知识、理解、应用与分析。虽然预测题目有风险，但某些模块——如 CPU 架构、二进制算术、网络协议和算法追踪——每轮考试都会出现。你应在考前至少完成四次限时全真模拟。

2. Data Representation & Number Systems | 数据表示与数制

Binary, hexadecimal, BCD and two’s complement form the core of early Paper 1 questions. Past papers show consistent tasks: convert positive and negative integers into two’s complement 8-bit form, perform binary addition, and interpret overflow. The mantissa-and-exponent floating-point representation appears less often but must be mastered for the top band.

二进制、十六进制、BCD 和补码是 Paper 1 早期题目的核心。历年考题中持续出现的任务包括：将正整数和负整数转换为 8 位补码形式、执行二进制加法并判断溢出。尾数和指数形式的浮点表示出现频率较低，但想要获得高分仍须掌握。

When two’s complement subtraction is required, the exam expects you to show the two’s complement of the subtrahend and then perform addition. Always check for overflow by comparing the carry into and out of the most significant bit. A common trap is forgetting to state whether overflow has occurred and linking that to the word length.

当需要进行补码减法时，考试期望你展示减数的补码，然后执行加法。务必通过比较最高位的进位人和进位出来检查溢出。一个常见陷阱是忘记说明是否发生溢出，并将其与字长联系起来。

For binary-coded decimal (BCD), past papers ask you to explain why BCD is used in digital clocks and financial systems. The answer revolves around the precise decimal representation without the rounding errors that occur in pure binary.

对于 BCD，往届真题会问为何数字钟表和金融系统使用 BCD。答案围绕精确的十进制表示，避免了纯二进制中的舍入误差。

Conversion	Example
-71 to 8-bit two’s complement	01000111 -> 10111000 + 1 = 10111001
BCD 94 to binary	1001 0100 BCD = 94 decimal = 01011110₂

3. Communication & Internet Technologies | 通信与互联网技术

This topic blends hardware transmission media with protocol stacks. Candidates must compare twisted pair, coaxial and fibre optic cables in terms of bandwidth, attenuation and interference. Fibre optic cable uses total internal reflection of light to transmit data with minimal signal loss over long distances, which is the premium answer examiners look for.

该主题融合了硬件传输介质与协议栈。考生必须对比双绞线、同轴电缆和光纤的带宽、衰减和抗干扰能力。光纤利用光的全内反射传输数据，在长距离上信号损耗极小，这是考官期望的高分答案。

On the protocol side, the TCP/IP four-layer model is examined almost every series. Be ready to explain the role of each layer: Application (HTTP, FTP, SMTP), Transport (TCP/UDP port addressing, segmenting), Internet (IP addressing, routing) and Network Access (MAC, frame). A high-mark question may ask you to describe how an email travels through the layers, combining SMTP at the Application layer with TCP segmentation and IP routing.

在协议方面，TCP/IP 四层模型几乎每轮都考。你要能解释各层的角色：应用层（HTTP、FTP、SMTP）、传输层（TCP/UDP 端口寻址、分段）、网络层（IP 寻址、路由）和网络接入层（MAC、帧）。高分题可能要求描述电子邮件如何逐层传输，结合应用层的 SMTP、TCP 分段和 IP 路由。

Mark schemes demand precise terms: ‘segments’ for TCP, ‘packets’ for IP, ‘frames’ for data link layer. Using the wrong term costs marks. When asked about packet switching, always mention that each packet may take a different route and that routers use the destination IP address to forward the packet to the next hop.

评分标准要求精确使用术语：TCP 用“segments”（段），IP 用“packets”（包），数据链路层用“frames”（帧）。术语错误会失分。当被问及分组交换时，一定要提到每个分组可能走不同的路径，路由器根据目的 IP 地址将分组转发到下一跳。

4. Processor Fundamentals | 处理器基础

The Von Neumann architecture and the fetch-decode-execute cycle are perennial favourites. A 6-mark question typically asks you to describe the cycle step-by-step, naming the registers involved: Program Counter (PC), Memory Address Register (MAR), Memory Data Register (MDR), Current Instruction Register (CIR) and Accumulator (ACC). The sequence must be clearly linked to the system buses: address, data, and control.

冯·诺依曼架构与取指-解码-执行周期是常考内容。一道 6 分的题目通常要求你逐步描述该周期，并指出所涉及的寄存器：程序计数器（PC）、内存地址寄存器（MAR）、内存数据寄存器（MDR）、当前指令寄存器（CIR）和累加器（ACC）。必须将每一步与系统总线（地址、数据、控制）明确联系起来。

The mark for the fetch phase is awarded when you state that the content of the PC is copied to the MAR and the address bus sends the address to memory, then the instruction is transferred via the data bus into the MDR and copied to the CIR. The PC is incremented either during or after the fetch. Decode and execute vary by instruction – be ready to give examples such as ADD or LOAD.

当你写出将 PC 的内容复制到 MAR、地址总线将地址发送到内存、然后指令经数据总线进入 MDR 并复制到 CIR 时，取指阶段的分数就拿到了。PC 在取指期间或之后递增。解码和执行阶段因指令而异——要做好准备举例说明 ADD 或 LOAD 等指令。

Interrupt handling is a higher-level concept tested through end-of-paper essays. The key is to describe saving the current context (registers) onto the stack, loading the Interrupt Service Routine (ISR) address into the PC, and restoring the context after the ISR completes. Always mention that the interrupt register is checked at the end of each FDE cycle.

中断处理是通过卷末论述题考查的更高阶概念。关键在于描述将当前上下文（寄存器）保存到堆栈，将中断服务例程（ISR）的地址加载到 PC，并在 ISR 完成后恢复上下文。别忘了提到每次 FDE 周期结束时都会检查中断寄存器。

5. System Software & Operating Systems | 系统软件与操作系统

Past papers ask you to distinguish between application software and system software, with operating systems, compilers, linkers and device drivers falling into the latter category. A common task is to explain how a high-level language program becomes executable code: pre-processing, compilation into assembly, assembly into machine code, and linking to library files. The loader then places the executable in memory.

往年试卷会要求你区分应用软件和系统软件，操作系统、编译器、链接器和设备驱动程序属于后者。一个常见任务是解释高级语言程序如何变成可执行代码：预处理、编译成汇编、汇编成机器码，再链接到库文件。然后加载器将可执行文件放入内存。

When an OS manages memory, marks are earned for describing paging, segmentation, and virtual memory. Paging divides memory into fixed-size blocks, which reduces external fragmentation. Virtual memory uses a portion of the hard disk as an extension of RAM; you must link this to the concept of swapping. Always note that excessive paging leads to disk thrashing and degraded performance.

当操作系统管理内存时，描述分页、分段和虚拟内存可以得分。分页将内存划分为固定大小的块，减少了外部碎片。虚拟内存使用硬盘的一部分作为 RAM 的扩展；你必须将此与交换的概念联系起来。务必指出过多的分页会导致磁盘颠簸和性能下降。

6. Security, Privacy & Ethics | 安全、隐私与伦理

Security threats and countermeasures appear in Paper 1 almost every year. Malware types – virus, worm, Trojan horse, spyware, ransomware – must be defined with their distinguishing feature. A virus attaches to a host file, a worm self-replicates across networks, a Trojan disguises as legitimate software. Phishing, pharming and denial-of-service attacks are the next layer of network threats you must be able to distinguish.

安全威胁与对策几乎每年都出现在 Paper 1 中。你需要定义各类恶意软件——病毒、蠕虫、特洛伊木马、间谍软件、勒索软件——并说明各自的特征：病毒附着在宿主文件上，蠕虫在网络中自我复制，特洛伊木马伪装成合法软件。钓鱼攻击、域欺骗和拒绝服务攻击是另一层你必须能够区分的网络威胁。

Data validation and verification are separate topics that candidates often confuse. Validation checks if data is sensible (range check, type check, format check) using algorithms like check digits. Verification ensures data has been entered correctly, e.g. double entry or visual check. The Luhn algorithm for credit card numbers or the ISBN-13 check digit calculation may be required as a worked example.

数据验证和校验是考生经常混淆的不同主题。验证检查数据是否合理（范围检查、类型检查、格式检查），可使用校验位等算法。校验则确保数据录入正确，例如双重录入或目视检查。信用卡号的 Luhn 算法或 ISBN-13 的校验位计算都可能被要求作为示例演示。

Ethics questions are threshold-level marks: you need to discuss data privacy, copyright, and the digital divide. Use the standard framework: state the issue, give a real-world example, and suggest a balanced solution. For example, when asked about smart home devices, mention the benefits of convenience vs. the risks of data collection without consent.

伦理问题属于基准分：你需要讨论数据隐私、版权和数字鸿沟。使用标准框架：陈述问题、给出真实案例、提出一个平衡的解决方案。例如，当被问及智能家居设备时，既要提到便利性的好处，也要说明未经同意收集数据的风险。

7. Algorithm Design & Problem-Solving | 算法设计与问题求解

Paper 2 pseudocode tasks mirror textbook patterns: linear search, binary search, bubble sort, insertion sort, and array manipulation. Past marks show that correctly initialising counters and loops is half the battle. A for loop must have a clearly defined start, end, and step; a while loop must ensure the condition eventually becomes false.

Paper 2 伪代码任务反映了教材中的典型模式：线性搜索、二分搜索、冒泡排序、插入排序和数组操作。以往评分显示，正确初始化计数器和循环就成功了一半。for 循环必须有明确定义的起点、终点和步长；while 循环必须确保条件最终会变为假。

Trace tables are systematically tested. You receive a pseudocode algorithm and a partially filled trace table. The most marks are lost by skipping a row or misjudging when a variable updates. Practise at least ten trace-table questions from past papers; always cross-check each row against the loop condition.

追踪表会系统性地考查。你会拿到伪代码算法和一张部分填充的追踪表。最常见的失分是漏写一行或误判变量更新时机。请至少练习十道来自历年真题的追踪表题目，并始终将每一行与循环条件进行交叉检查。

Abstract Data Types (ADTs) such as stack and queue are examined either in pseudocode or by describing their operations: push, pop, isEmpty, isFull for a stack; enqueue, dequeue for a queue. Be prepared to use an array and a pointer (top or front/rear) in a pseudocode implementation.

抽象数据类型（ADT）如栈和队列可能以伪代码形式考查，或要求描述其操作：栈的 push、pop、isEmpty、isFull；队列的 enqueue、dequeue。你需要准备好用数组和指针（top 或 front/rear）编写伪代码实现。

8. Programming Concepts & Data Structures | 编程概念与数据结构

Beyond the basic algorithms, Cambridge examiners like to test 2D array manipulation and file handling in pseudocode. A typical question might read a comma-separated text file into a 2D array and then perform a search or sort. Open, read, write and close operations on text files must be demonstrated with proper syntax: OPENFILE "data.txt" FOR READ, READFILE "data.txt", Line.

除了基础算法，剑桥考官喜欢在伪代码中考查二维数组操作和文件处理。典型题目可能是将一个逗号分隔的文本文件读入二维数组，然后执行搜索或排序。必须用正确语法展示文本文件的打开、读取、写入和关闭操作：OPENFILE“data.txt”FOR READ、READFILE“data.txt”, Line。

Structured programming principles – modularity, meaningful identifiers, comments, parameters and return values – are rewarded both in code-writing and in long-answer questions. When asked to evaluate a piece of code, comment on its readability, use of functions, and whether it avoids global variables.

结构化编程原则——模块化、有意义的标识符、注释、参数和返回值——在编写代码和长答题中都能获得奖励。当被要求评估一段代码时，要评论其可读性、函数的使用，以及是否避免了全局变量。

9. Databases & SQL | 数据库与 SQL

Database questions typically show a table structure and ask you to write SQL queries. The tested subset is small: SELECT, FROM, WHERE, ORDER BY, GROUP BY, INNER JOIN, COUNT, SUM, AVG. Sub-queries appear rarely but can distinguish top candidates. Always format SQL keywords in uppercase to satisfy the mark scheme.

数据库题目通常会给出表结构并让你编写 SQL 查询。考查的子集很小：SELECT、FROM、WHERE、ORDER BY、GROUP BY、INNER JOIN、COUNT、SUM、AVG。子查询很少出现，但能让高分考生脱颖而出。请始终将 SQL 关键字大写，以满足评分标准。

Normalisation to Third Normal Form (3NF) is tested by giving you unnormalised data and asking to produce relations. The process: remove repeating groups to 1NF, remove partial dependencies to 2NF, remove transitive dependencies to 3NF. The final schema must be drawn with primary keys underlined and foreign keys indicated.

第三范式（3NF）的规范化考查方式是给出未规范化的数据，要求生成关系。过程为：移除重复组至 1NF，移除部分依赖至 2NF，移除传递依赖至 3NF。最终模式必须绘制，主键下划线标示，外键予以注明。

10. Common Mistakes & Examiner Tips | 常见错误与考官建议

Examiner reports highlight the same weaknesses year after year. In binary addition, candidates forget to indicate overflow. In the FDE cycle, they omit to mention the increment of the PC. In pseudocode, they use = both for assignment and comparison – assignment should be ←, comparison =. Also, declaring a variable as a string but comparing it to an integer loses all logic marks.

考官报告年复一年地指出同样的薄弱环节。在二进制加法中，考生忘记指出溢出。在 FDE 周期中，他们忘记提及 PC 的递增。在伪代码中，他们把 = 同时用于赋值和比较——赋值应使用 ←，比较用 =。此外，将变量声明为字符串却与整数比较，会失去所有逻辑分。

For longer essay questions, structure your answer with a brief introduction, two or three well-developed points, and a conclusion that answers the exact question. Use technical vocabulary precisely. If a question says ‘describe’, give facts and processes; if it says ‘explain’, add reasons and impacts; ‘evaluate’ requires weighing advantages against disadvantages before a final judgement.

对于较长的论述题，要用简短的引言、两到三个充分展开的要点以及一个精准回应问题的结论来组织答案。严谨使用技术词汇。如果题目说“describe（描述）”，给出事实和过程；如果说“explain（解释）”，增加原因和影响；而“evaluate（评价）”则要求在做出最终判断之前权衡优缺点。

Time management is critical. Allocate one minute per mark for Paper 1 and 1.6 minutes per mark for Paper 2. Leave 10 minutes at the end to review trace tables and SQL queries – those sections have the highest catch-able errors. Practise past papers until you can predict the next logical step before reading the mark scheme.

时间管理至关重要。Paper 1 按每分钟一分、Paper 2 按每 1.6 分钟一分来分配时间。最后留出 10 分钟检查追踪表和 SQL 查询——这些部分最易发现错误。反复练习历年真题，直到你能在查阅评分方案之前就预判出下一个逻辑步骤。

Published by TutorHao | Computer Science Revision Series | aleveler.com

更多咨询请联系16621398022（同微信）