PG Engineering · IIT Madras (GATE 2027), on behalf of the National Coordination Board — GATE, MoE
GATE — Graduate Aptitude Test in Engineering — Computer Science & Information Technology (CS)
The complete GATE blueprint: tier-wise pattern, topic weightage from previous-year analyses, and free chapter-by-chapter study material.
65 questions, 100 marks, 180 minutesMCQ only carries negative marking — MSQ and NAT do not3 years score validity Free · tutor-verified
GATE at a glance
Conducted byIIT Madras (GATE 2027), on behalf of the National Coordination Board — GATE, MoE
CycleOnce a year, across three weekends in February
ModeComputer-based test — MCQ, MSQ (multiple select) and NAT (numerical answer, typed) questions in one paper
EligibilityCurrently studying in the 3rd year or higher of any undergraduate degree, or already holding a degree in Engineering, Technology, Science, Commerce, Arts or Architecture. There is no age limit and no restriction on the number of attempts.
LanguagesEnglish only
Qualifies you forM.Tech/MS/PhD admission at IITs, IISc, NITs and IIITs (with MHRD stipend), plus direct recruitment at PSUs including BHEL, IOCL, ONGC, NTPC, PGCIL, GAIL and BARC
What a GATE CS score opens up
M.Tech / MS by Research at IITs and IIScPhD admission with a monthly research stipendPSU recruitment (BHEL, IOCL, ONGC, PGCIL, GAIL)Scientist posts via BARC, ISRO and DRDO screeningJunior Research Fellow positions at national labsFaculty eligibility at several state technical universities
⚡ What changed recently
- GATE 2027 is organised by IIT Madras; GATE 2026 was organised by IIT Guwahati. The organising institute rotates annually among the IITs and IISc, and it does not change the syllabus.
- Negative marking applies to MCQs ONLY — 1/3 mark for a wrong 1-mark MCQ and 2/3 for a wrong 2-mark MCQ. MSQ (multiple select) and NAT (numerical answer) questions carry no negative marking at all, which makes attempting every NAT and MSQ strictly correct strategy.
- The paper is one freely navigable 180-minute window with no sectional time locks and no forced ordering — you may move between General Aptitude, Engineering Mathematics and core CS in any order, unlike exams that lock groups.
- A GATE score is valid for three years from the date of announcement, so a strong score can be used across three admission and recruitment cycles.
- Candidates may appear in two papers from a permitted combination list, which for CS candidates most commonly pairs CS with Data Science & Artificial Intelligence (DA).
How GATE CS selection works
Know what each stage is for before you spend a single hour preparing.
PRODUCES THE GATE SCORE AND ALL INDIA RANK
GATE CS — single computer-based paper
65 questions, 100 marks, 180 minutes. 30 questions carry 1 mark and 35 carry 2 marks. Negative marking applies to MCQs only. A virtual calculator is provided on screen; personal calculators are barred.
CONVERTS RAW MARKS TO A GATE SCORE
Score normalisation and result
Where a paper runs in multiple sessions, raw marks are normalised across sessions before the GATE score is computed, so rank depends on normalised performance rather than raw marks in isolation. Qualifying marks are category-wise and are announced with the result.
WHERE THE SCORE IS ACTUALLY USED
Admission or PSU recruitment
Institutes conduct their own counselling (COAP for the IITs) using the GATE score, sometimes with an interview or written test. PSUs publish separate GATE-score cutoffs and run their own interview and document verification rounds.
The exam pattern, tier by tier
Marks, timing and negative marking exactly as per the official notification.
GATE CS — General Aptitude, Engineering Mathematics and core Computer Science
Single paper; the score it produces is what admissions and PSUs use100 marks · 65 Q
180 min (240 min for PwD candidates with scribe eligibility), single freely navigable window · MCQ only: −1/3 for a wrong 1-mark question, −2/3 for a wrong 2-mark question. MSQ and NAT carry no negative marking.
| Section | Questions | Marks | Time |
|---|---|---|---|
| General Aptitude | ~10 | 15 | — min |
| Engineering Mathematics | ~9 | ~13 | — min |
| Core Computer Science | ~46 | ~72 | — min |
💡 Question-type marks are fixed at 30 one-mark and 35 two-mark questions. General Aptitude is a fixed 15 marks in every GATE paper. The split of the remaining 85 marks across Engineering Mathematics and the core subjects varies year to year.
Where GATE CS's 100 marks come from
One paper, one 180-minute window, freely navigable — there are no locked sections. General Aptitude is a fixed 15 marks in every GATE paper; Engineering Mathematics contributes about 13; the remaining ~72 marks are core Computer Science. Question-type marks are fixed (30 one-mark and 35 two-mark questions), but the split across core subjects varies year to year, so the per-subject figures below are estimates from past-paper analysis. Click any block to open its chapters:
General Aptitude15 marks · 15%
A fixed 15 marks in every GATE paper — the only block whose weight is guaranteed. Verbal and quantitative aptitude dominate; the questions are deliberately below engineering level and are the cheapest marks on the paper.
Engineering Mathematics13 marks · 13%
About 13 marks, with Discrete Mathematics the heaviest of the four chapters. Probability and Linear Algebra questions are usually short and computational, which makes them high-yield per minute.
Programming & Data Structures11 marks · 11%
The heaviest core subject. C output questions and pointer-based memory reasoning recur every year, and data structure questions almost always come down to counting or a traversal argument.
Operating Systems9 marks · 9%
Concurrency and deadlock is the single densest area, followed by scheduling numericals and virtual memory arithmetic. Almost every question is computational rather than descriptive.
Algorithms9 marks · 9%
Recurrences, sorting bounds and graph algorithms recur annually. Questions test whether you can bound growth and justify a greedy or dynamic-programming choice, not whether you can code.
Computer Networks9 marks · 9%
Delay arithmetic, sliding-window efficiency, subnetting and TCP congestion window traces are the reliable numericals. The four delay components must be kept separate to score here at all.
Databases8 marks · 8%
Normal forms and candidate keys, B+ tree order computation, and serializability testing via precedence graphs. Nearly every question is mechanical once the right procedure is chosen.
Theory of Computation8 marks · 8%
Closure properties, pumping lemma arguments and decidability classification. This is the most conceptual subject in the paper and rewards precise definitions over memorised results.
Computer Organization & Architecture7 marks · 7%
Pipelining speedup, cache hit-rate arithmetic and addressing modes. Memory hierarchy alone is usually the largest single contributor within this subject.
Digital Logic6 marks · 6%
Number representation, K-map minimisation and sequential circuit analysis. Short, self-contained questions that are among the fastest marks in the core section.
Compiler Design5 marks · 5%
The lightest core subject, and the most predictable: parser classification, first and follow sets, and data-flow analysis classification account for most of what is asked.
Subjects in GATE
Chapter-by-chapter study material built to the exam's own blueprint.
Topic-wise weightage — what to study first
Question counts per topic, distilled from previous-year paper analyses. Click any topic with a link to open its full chapter.
| Topic | Marks | Priority |
|---|---|---|
| Verbal Aptitude | ~4 | Very high |
| Quantitative Aptitude | ~5 | Very high |
| Analytical Aptitude | ~4 | High |
| Spatial Aptitude | ~2 | Medium |
🎯 General Aptitude is the only section whose weight is fixed by regulation rather than varying with the year, which makes it the most reliably plannable 15 marks on the paper. The questions sit deliberately below engineering level: verbal aptitude tests whether you decide by grammatical structure or by the text rather than by what sounds right, and quantitative aptitude rewards finding the invariant instead of grinding through arithmetic. Analytical aptitude questions are almost always constraints plus elimination, and spatial aptitude reduces to tracking one feature through a transformation rather than visualising the whole object. Candidates routinely under-prepare this section on the assumption that it is easy, then lose marks to careless reading — the marks are cheap but not free.
| Topic | Marks | Priority |
|---|---|---|
| Discrete Mathematics | ~5 | Very high |
| Linear Algebra | ~3 | High |
| Calculus | ~2 | Medium |
| Probability and Statistics | ~3 | High |
🎯 Engineering Mathematics is where the highest marks-per-minute sit, because its questions are short and self-contained. Discrete Mathematics is the heaviest chapter and the one that feeds directly into Theory of Computation and Algorithms — proof by counterexample, counting arguments and graph properties all reappear there. Linear Algebra questions almost always reduce to rank, so establishing rank first answers most of them without further computation. Calculus questions stay at limits, continuity, differentiability and definite integrals, and the marks usually live in the hypotheses of a theorem rather than in its conclusion. Probability questions are won by naming the sample space explicitly before doing any arithmetic, which is exactly where careless candidates go wrong.
| Topic | Marks | Priority |
|---|---|---|
| Number Representation & Computer Arithmetic | ~2 | Very high |
| Boolean Algebra & K-maps | ~2 | High |
| Combinational Circuits | ~1 | Medium |
| Sequential Circuits | ~1-2 | High |
🎯 Digital Logic questions are among the fastest in the core section because each is self-contained: nothing depends on a long chain of earlier reasoning. Number representation is the chapter that quietly decides several marks elsewhere too, since two's complement overflow and floating-point rounding reappear in Computer Organization. Boolean minimisation is a covering problem, and recognising it as such is what prevents the common error of stopping at a locally small expression. Combinational circuit questions always reduce to a truth table, however the circuit is drawn, and sequential circuit questions reduce to asking what history the state bits must remember. Attempt this subject early in the paper while your arithmetic is still careful.
| Topic | Marks | Priority |
|---|---|---|
| Machine Instructions & Addressing Modes | ~1 | Medium |
| ALU, Data-path & Control Unit | ~1 | Medium |
| Instruction Pipelining & Hazards | ~2 | Very high |
| Memory Hierarchy: Cache & Main Memory | ~2-3 | Very high |
| I/O Interface: Interrupt & DMA Mode | ~1 | Medium |
🎯 This subject is almost entirely numerical, and the numericals fall into three families that recur every year: pipelining speedup with hazards, cache hit-rate and average access time, and effective address computation under a named addressing mode. Memory hierarchy alone usually contributes the most marks, and its questions turn on splitting an address into tag, index and offset fields correctly — get that split wrong and every subsequent number is wrong. Pipelining questions are decided by identifying which overlap is unsafe rather than by the speedup formula itself. The I/O chapter is smaller but reliably examined through the interrupt-versus-DMA comparison, which is really a question about who waits and who moves the bytes.
| Topic | Marks | Priority |
|---|---|---|
| Programming in C & Recursion | ~3 | Very high |
| Arrays, Stacks, Queues & Linked Lists | ~3 | Very high |
| Trees & Binary Search Trees | ~3 | Very high |
| Binary Heaps & Graphs as Data Structures | ~2 | High |
🎯 This is the heaviest core subject and the one where careful, mechanical work pays most. C questions are output-prediction problems, and the only reliable method is to draw the memory rather than reason about the code in your head, because C is a thin layer over addresses and every trick question exploits that. Recursion questions are answered by writing the call tree, not by tracing mentally. For data structures, the organising question is always where insertions and deletions are allowed, which is what distinguishes a stack from a queue from a deque. Tree questions come down to counting arguments or traversals, and heap and graph questions come down to maintaining only what the queries actually need. Attempt the pointer-heavy questions when you are fresh.
| Topic | Marks | Priority |
|---|---|---|
| Asymptotic Complexity & Recurrences | ~2 | Very high |
| Searching, Sorting & Hashing | ~2-3 | Very high |
| Algorithm Design: Greedy, Divide & Conquer, Dynamic Programming | ~2 | High |
| Graph Algorithms | ~2-3 | Very high |
🎯 GATE tests whether you can bound growth and justify a choice, never whether you can write working code. Asymptotic questions are about growth rate rather than speed, and the master theorem plus a recursion tree covers almost every recurrence that appears. Sorting and hashing questions rest on two bounds worth internalising: the information-theoretic lower bound on comparison sorting, and the load factor's control of hashing performance. Design technique questions ask how subproblems relate — overlapping means dynamic programming, independent means divide and conquer, and a provable exchange argument means greedy. Graph algorithm questions are traversals with bookkeeping, and the distinguishing question is when the algorithm dares to finalise a vertex, which is exactly why Dijkstra fails on negative edges.
| Topic | Marks | Priority |
|---|---|---|
| Regular Expressions & Finite Automata | ~2-3 | Very high |
| Context-Free Grammars & Pushdown Automata | ~2 | High |
| Regular & Context-Free Languages: Closure and the Pumping Lemma | ~2 | High |
| Turing Machines & Undecidability | ~2 | Very high |
🎯 Theory of Computation rewards precise definitions more than any other subject here, because every question turns on a distinction that a loose statement erases. Finite automata questions come down to how many situations a machine must tell apart, which is what the Myhill-Nerode argument formalises and what makes minimisation and non-regularity proofs the same idea. Pushdown automata exist because a stack is exactly the memory that nesting requires, so recognising nesting in a language usually settles its class immediately. The pumping lemma is a game against an adversary and must be argued in the right order, which is where most marks are lost. Undecidability questions rest on self-reference making diagonalisation available, and reduction is the standard tool.
| Topic | Marks | Priority |
|---|---|---|
| Lexical Analysis, Parsing & Syntax-Directed Translation | ~2 | Very high |
| Intermediate Code Generation | ~1 | Medium |
| Runtime Environments | ~1 | Medium |
| Local Optimisation & Data Flow Analysis | ~1-2 | High |
🎯 Compiler Design is the smallest core subject but has the highest ratio of predictable questions to syllabus, which makes it excellent value late in a preparation cycle. Parsing questions are settled by two facts: how much lookahead is available and whether the tree is built top-down or bottom-up, which together define every parser class. First and follow set computation and LL(1) conflict detection appear almost annually. Intermediate code generation questions are about turning an m-by-n problem into an m-plus-n one, and runtime environment questions ask where storage lives and how long it lasts. Data flow analysis is answered by classifying the analysis on two axes, direction and meet operator, which is close to a free mark once memorised.
| Topic | Marks | Priority |
|---|---|---|
| Processes, Threads, System Calls & IPC | ~1-2 | High |
| CPU & I/O Scheduling | ~2 | Very high |
| Concurrency, Synchronization & Deadlock | ~2-3 | Very high |
| Memory Management & Virtual Memory | ~2 | Very high |
| File Systems | ~1 | Medium |
🎯 Operating Systems questions are computational far more often than descriptive, and each chapter has a signature numerical. Process questions are fork-counting exercises plus the shared-versus-private table for threads, both of which are pure bookkeeping. Scheduling questions are Gantt charts followed by an arithmetic mean, where the marks are lost to arrival times and tie-breaking rather than to conceptual confusion. Concurrency is the densest area, covering semaphore ordering, the banker's algorithm and the minimum-resource formula. Memory management is address-field splitting and effective access time, and Belady's anomaly is the recurring conceptual question. File systems reduce to inode arithmetic: maximum file size and the number of accesses to reach a given byte.
| Topic | Marks | Priority |
|---|---|---|
| ER Model & Relational Model | ~1 | Medium |
| Relational Algebra, Tuple Calculus & SQL | ~1-2 | High |
| Integrity Constraints & Normal Forms | ~2 | Very high |
| File Organization & Indexing (B and B+ Trees) | ~1-2 | Very high |
| Transactions & Concurrency Control | ~2 | Very high |
🎯 Databases is the most procedural subject in the paper: almost every question has a fixed method, and the marks go to whoever applies it without slipping. Candidate key finding and highest-normal-form identification appear almost every year, and both depend on computing attribute closures correctly and finding every candidate key rather than just one. SQL questions concentrate on where theory and practice diverge, namely set versus multiset semantics and three-valued logic with nulls, which is where the NOT IN trap lives. B-plus tree order computation is a near-certain numerical and hinges on computing the internal and leaf orders separately. Serializability is decided by a precedence graph cycle check, which takes a minute and cannot go wrong if the conflicting pairs are listed item by item.
🎯 Computer Networks marks come almost entirely from arithmetic, and the single largest cause of lost marks is merging transmission delay with propagation delay, which are independent quantities. Once the four delay components are kept separate, the store-and-forward pipelining formula and the bandwidth-delay product follow directly. Data link layer questions are dominated by sliding window efficiency, which is entirely a function of the propagation-to-transmission ratio, plus the CSMA/CD minimum frame size derivation. Network layer questions are subnetting, CIDR aggregation, longest prefix match and fragment offsets, all mechanical. Transport layer questions are congestion window traces, where the threshold must be updated explicitly at every loss event, and the application layer contributes DNS message counts and HTTP round-trip comparisons.
GATE resources
Everything you need to prepare.
Need structured GATE guidance?
Combine our free study material with a live course or 1-on-1 tutoring.
FAQs
Frequently asked questions
65 questions for 100 marks in a single 180-minute computer-based test. 30 questions carry 1 mark and 35 carry 2 marks. There are no sectional time locks — the whole paper is one freely navigable window, so you may move between General Aptitude, Engineering Mathematics and the core subjects in any order.
Negative marking applies to multiple-choice questions ONLY: −1/3 for a wrong 1-mark MCQ and −2/3 for a wrong 2-mark MCQ. Multiple-select (MSQ) and numerical-answer (NAT) questions carry no negative marking at all. That asymmetry has a direct consequence — you should attempt every NAT and MSQ question, since a wrong answer costs nothing, while MCQs should be guessed only after eliminating at least one option.
Only partly. General Aptitude is fixed at 15 marks in every GATE paper, and the 30-plus-35 split of 1-mark and 2-mark questions is fixed. Beyond that, the distribution across Engineering Mathematics and the ten core subjects varies year to year and is not published in advance. The per-subject figures shown here are estimates from past-paper analysis, useful for planning but not a guarantee for any single year.
GATE 2027 is organised by IIT Madras; GATE 2026 was organised by IIT Guwahati. The responsibility rotates annually among the IITs and IISc. The organising institute handles administration and paper-setting coordination, but the syllabus is set nationally and does not change with the rotation.
Three years from the date the result is announced. Within that window it can be used for M.Tech, MS by Research and PhD admission at the IITs, IISc, NITs and IIITs — with an MHRD stipend for admitted students — and for direct recruitment at PSUs including BHEL, IOCL, ONGC, PGCIL, GAIL and NTPC, each of which publishes its own GATE-score cutoff.
Anyone currently in the third year or higher of an undergraduate degree, or already holding a degree in Engineering, Technology, Science, Commerce, Arts or Architecture. There is no age limit and no cap on the number of attempts. Candidates may also appear in two papers from a permitted combination list, which for CS candidates most commonly pairs CS with Data Science & Artificial Intelligence.
General Aptitude first, because its 15 marks are guaranteed and the questions sit below engineering level. Then Engineering Mathematics, especially Discrete Mathematics, which feeds directly into Theory of Computation and Algorithms. Among core subjects, Programming & Data Structures is consistently the heaviest, followed by Operating Systems, Algorithms and Computer Networks at roughly 9 marks each. Compiler Design is the lightest but has the highest ratio of predictable questions to syllabus, which makes it good value late in a cycle.
Take the next step
GATE — find a tutor
Pair self-study with a tutor, a live course or a coaching centre.
