Back to Study Material GATE Exam

COA Previous Year Questions

Analyzing previous year GATE questions in Computer Organization and Architecture reveals consistent patterns that guide effective preparation. Over the pas...

COA Previous Year Questions

Patterns in GATE COA Questions from Past Years

Analyzing previous year GATE questions in Computer Organization and Architecture reveals consistent patterns that guide effective preparation. Over the past decade, pipeline CPI calculations, cache address decomposition, and memory hierarchy AMAT problems have appeared in almost every paper. Understanding these patterns helps you allocate study time optimally.

Pipeline questions from GATE 2019 and 2021 asked students to identify the number of clock cycles required to execute a given sequence of instructions with specified hazards. The trap many students fall into is forgetting that a stall bubble occupies a pipeline slot just like a real instruction. When the question says a hazard causes a two-cycle stall, two NOP-like bubbles are inserted, pausing the pipeline for two cycles at that point in the instruction stream.

Cache questions often provide a system with a specified cache size, block size, and associativity, then ask for the number of tag bits, index bits, or offset bits. GATE 2020 presented a 32 KB direct-mapped cache with 32-byte lines and asked for the tag field width given a 32-bit physical address. The solution: offset = log2(32) = 5 bits; index = log2(32768 divided by 32) = log2(1024) = 10 bits; tag = 32 minus 5 minus 10 = 17 bits.

Interrupt and I/O questions from GATE 2017 and 2018 computed the fraction of CPU time consumed by handling device interrupts. The approach is straightforward: determine how many interrupts occur per second and multiply by the CPU cycles consumed per interrupt (context save, ISR execution, context restore). Dividing by total CPU cycles per second gives the fraction.

Memory interleaving questions ask how many modules are needed to hide a given memory latency. If memory access takes 40 ns and the bus cycle is 10 ns, four interleaved modules ensure that by the time one module responds, the next has already been accessed for one cycle — though exact formulas vary, the principle is to match module count to the ratio of memory latency to bus cycle time.

Solving COA Numerical PYQs Step by Step

This section walks through the methodology for solving common numerical COA PYQs efficiently and without arithmetic errors. Developing a structured approach is more valuable than memorizing specific solutions, because GATE regularly modifies numbers while keeping the underlying problem structure identical.

Step one: identify the question category. Is it pipeline (CPI calculation), cache (address decomposition or AMAT), memory (effective access time with TLB), or I/O (transfer rate or CPU overhead)? Each category has a standard formula. Write it down immediately before substituting any numbers.

Step two: extract all given values and assign units. GATE questions sometimes mix nanoseconds and microseconds, or cycles and seconds. Convert everything to a common unit before computing. A frequent error is treating a 100 MHz clock as 100 ns per cycle instead of 10 ns per cycle.

Step three: apply the formula. For AMAT with a two-level cache: AMAT = 1 ns (L1 hit time) + 0.05 (L1 miss rate) times [10 ns (L2 hit time) + 0.2 (L2 miss rate) times 100 ns (main memory)]. Compute the innermost bracketed expression first: 10 + 0.2 times 100 = 10 + 20 = 30 ns. Then: 1 + 0.05 times 30 = 1 + 1.5 = 2.5 ns.

Step four: sanity-check your answer. If AMAT comes out larger than main memory access time, something is wrong — AMAT must always be between L1 hit time and main memory access time. If effective CPI is less than 1, you have an error.

Step five: eliminate wrong options logically. GATE multiple-choice options are usually spread across a reasonable range. If two options differ by exactly one pipeline stage, you likely missed a stall. If two options differ by a factor of two, you may have confused a two-way set-associative cache with a direct-mapped cache. Use option analysis to catch arithmetic mistakes under exam pressure.

High-Yield COA Topics Based on PYQ Frequency

Not all COA sub-topics carry equal weight in GATE historical papers. A frequency analysis of questions from GATE 2015 through 2024 reveals that roughly 40 percent of COA marks come from pipeline performance, 30 percent from cache and memory hierarchy, 15 percent from I/O and interrupts, and 15 percent from instruction formats and addressing modes.

Pipeline performance sub-topics with the highest frequency are: effective CPI computation with data and control hazards, total execution time for a sequence of instructions, and speedup comparison of pipelined versus non-pipelined execution. Questions in this cluster are almost guaranteed to appear in every paper, often as two-mark problems.

Cache and memory hierarchy questions most often test: direct-mapped versus set-associative cache address decomposition, AMAT calculation with one or two cache levels, replacement policy comparison for a given reference string (computing hit/miss counts for LRU versus FIFO), and write-back versus write-through traffic estimation. Expect at least one such question each year.

I/O questions typically accompany a scenario describing a peripheral device with known transfer rate and testing whether you can compute CPU utilization under programmed I/O, interrupt-driven I/O, or DMA. DMA questions sometimes combine with bus bandwidth calculations.

Instruction format questions are less frequent but high-value when they appear, often as one-mark questions asking for the number of addressable memory locations given a specific address field width. Addressing mode questions may ask you to trace the sequence of memory accesses for a given mode (direct, indirect, register-indirect) given register and memory contents.

Recommended last-week study plan: days one and two on pipeline with 20 practice problems, days three and four on cache with 15 problems, day five on I/O and interrupts, day six on instruction formats and addressing modes, day seven on timed full-paper simulation using recent GATE papers.