The CPU and Von Neumann Architecture
The Central Processing Unit (CPU) is the brain of the computer, comprising three core parts:
- ALU (Arithmetic Logic Unit): performs all arithmetic (+, -, x, /) and logical (AND, OR, comparison) operations.
- CU (Control Unit): directs and coordinates all operations; fetches, decodes and executes instructions but does NOT process data itself.
- Registers/MU: small high-speed temporary storage inside the CPU (e.g. Accumulator, PC, MAR, MBR).
Von Neumann architecture (stored-program concept) keeps both instructions and data in the SAME memory. The Machine Cycle = Fetch -> Decode -> Execute -> Store (FDES). CPU speed is measured in Hertz (GHz). Memory aid: ALU calculates, CU controls, Registers remember.
Motherboard, Buses and Ports
"Find the wrong term in the series" is one of the highest-scoring question types in RPF SI Reasoning — easy if you have a method, a coin-toss if you don't. This lesson walks you through that method using a clean worked example and the pattern-spotting habits a topper uses.
Definition: A number series is an ordered list of numbers generated by a fixed rule.
Definition: A wrong-term question gives you a series in which exactly one term breaks the rule and asks you to identify it (and sometimes give its correct replacement).
The given problem
Find the wrong term in: 4, 9, 19, 39, 80, 159.
How toppers approach it
There are essentially four rule-families to test, in order of frequency:
- Constant operation — add/subtract a fixed number (arithmetic progression) or multiply/divide by a fixed ratio (geometric progression).
- Linear recurrence — each term is a × previous + b. This is the most common trap family.
- Differences forming a pattern — first differences are themselves an arithmetic or geometric series.
- Mixed operations — alternate +k, ×k, square, cube, etc.
For wrong-term questions the linear recurrence family (rule 2) shows up surprisingly often, because it produces near-doubling sequences where one off-by-one slip is easy to hide.
Solving 4, 9, 19, 39, 80, 159
Solution:
Step 1: Look at the ratios. 9/4 = 2.25, 19/9 ≈ 2.11, 39/19 ≈ 2.05, 80/39 ≈ 2.05, 159/80 ≈ 1.99. The ratios are clustering around 2, but are slightly more than 2 at the start. This is the fingerprint of a recurrence like x_{n+1} = 2x_n + c.
Step 2: Test the rule x_{n+1} = 2 x_n + 1:
- Start with 4. Apply the rule: 2(4) + 1 = 9. ✓
- From 9: 2(9) + 1 = 19. ✓
- From 19: 2(19) + 1 = 39. ✓
- From 39: 2(39) + 1 = 79. ✗ But the series says 80.
- From the correct value 79: 2(79) + 1 = 159. ✓ And the series says 159. ✓
Step 3: So the rule generates 4, 9, 19, 39, 79, 159 — and the given series differs from this only at one position. The wrong term is 80; it should be 79.
Conclusion: The wrong term is 80, and the correct replacement is 79.
The critical verification step
Notice what happened at the very end of Step 2: even after spotting that 80 didn't fit, we did one more check — we asked "if 79 was the real value, does the rule still produce 159 from it?" It does. That single extra calculation is what separates a clean answer from a wrong one.
Why is this step so important? Because the same data could in principle be generated by a different rule that happens to fit five of the six terms differently. If 80 had been the genuine wrong term and 159 also failed to fit the rule, then maybe the rule itself was misidentified and you should re-test other rules. By confirming that 159 = 2(79) + 1, we prove that 80 is the only anomaly, which is exactly the condition the question demands.
Why it matters
Why it matters: RPF SI and similar SSC/RRB exams have a tight clock (about 45 seconds per Reasoning question). Wrong-term questions reward a structured approach over flair: pick a rule, project it forward, look for the single break. If you start by inspection alone, you waste minutes second-guessing yourself. With this method, three or four attempts at a rule will crack almost any series.
The rule-family checklist (apply in order)
When you face an unknown series, run these tests roughly in order:
- Compute first differences: 5, 10, 20, 41, 79. Are they constant? Doubling? Arithmetic on their own?
- Compute ratios: 2.25, 2.11, 2.05, 2.05, 1.99. Are they tending to a clean integer? That hints at a recurrence x_{n+1} = k x_n + c. Solve for k and c using the first two terms.
- Check x² + 1 or x² − 1 style rules: 2² + 1 = 5? 3² = 9? Not quite, but worth a glance.
- Check x_{n+1} = x_n + (x_n − x_{n−1}) × something — second differences.
- Check alternating patterns by separating odd-indexed and even-indexed terms.
For our problem, the ratio test in (2) instantly suggested 2x+1.
Real-world example
Real-world example: Number series of this kind are basically the same logic that fraud-detection systems use when checking transaction patterns — most entries follow a rule (a customer's normal spending pattern), and one breaks it (a fraudulent charge). The mental skill of "find the rule, project it, flag the outlier" is exactly what your reasoning section is training. Indian fintech firms test this skill on aptitude rounds for the same reason.
A second worked check — different rule, same method
Question: Find the wrong term: 7, 15, 31, 63, 128, 255.
Solution:
Step 1: Ratios near 2 again. Try 2x + 1: from 7 → 15 ✓; 15 → 31 ✓; 31 → 63 ✓; 63 → 127 ✗ (given 128); 127 → 255 ✓.
Step 2: 128 is wrong; should be 127.
Conclusion: Wrong term = 128; correct = 127.
Same template, same answer style. Once you have the algorithm, you can do these in under 40 seconds.
Common misconception
Common misconception: "If a term doesn't fit, just say it's wrong and move on." Wrong. You must also verify that every later term fits the rule again when generated from the corrected value. Otherwise you may have misidentified the rule. In our original problem, if 80 had been the actual rule's output (say the rule was x_{n+1} = 2x_n + 2), then 159 wouldn't fit either, and the question would be ambiguous. By checking forward we eliminate that possibility and lock in the answer.
Another trap: confusing the wrong term with the next term. The question is about an error inside the series, not what comes after the series. Read carefully — RPF SI options usually include both, hoping you misread.
A third trap: assuming the rule must be "elegant". Real exam rules sometimes involve squares plus a small constant or alternating operations. Don't reject a rule just because it isn't a one-line magic formula; if it generates the correct value at every other position, it is the rule.
| Step | What you do | What it tells you |
|---|---|---|
| Differences | Compute t_{n+1} − t_n | Spots arithmetic / quadratic patterns |
| Ratios | Compute t_{n+1} / t_n | Spots geometric or linear-recurrence patterns |
| Project rule | Apply hypothesised rule forward | Finds the single mismatch |
| Verify after error | Generate next term from correct value | Confirms the rule, not a coincidence |
- ✓- Wrong-term questions ALWAYS have exactly one anomaly — your job is to confirm it is one.
- ✓- Pick a candidate rule using first differences, ratios, or recurrence form.
- ✓- Apply the rule from the first term forward, term by term.
- ✓- When a mismatch appears, generate the next term from the correct projected value, not from the wrong one — and check that later terms fit.
- ✓- Common recurrence forms in this question type: x²+1, 2x+1, 2x−1, 3x+2, alternating +/×.
- ✓- A rule that fits all but one position is the right rule.
"Project, Spot, Verify Forward" — the three-step mantra. Project the rule forward; spot the single break; verify the rule still holds after the break by feeding in the corrected value.
- ✓- The rule was x_{n+1} = 2 x_n + 1.
- ✓- 4 → 9 → 19 → 39 → 79 → 159 — the series puts 80 where 79 should be.
- ✓- Always project the rule forward past the suspected error to confirm it is the only one.
- ✓- The wrong term is 80; the correct term is 79.
Worked Example: Address Bus and Memory Capacity
Q: If a CPU has a 16-bit address bus, what is the maximum memory it can directly address?
Formula: Addressable memory = 2^(address bus width) locations.
Step 1: Width = 16 bits, so 2^16 = 65,536 locations = 64 KB (since 2^10 = 1 KB, 2^16 = 2^6 x 2^10 = 64 KB).
Answer: 64 KB.
Quick reference for the exam: 2^10=1K, 2^20=1M, 2^30=1G. A 20-bit bus = 2^20 = 1 MB; a 32-bit bus = 2^32 = 4 GB. Speed trick: subtract 10 from the exponent per K-step. 2^16 -> (16-10)=6, so 2^6=64 K = 64 KB. This 'subtract 10' trick saves time in DI-style memory questions.
Computer Architecture & Components — revision notes (IBPS PO)
Architecture questions in IBPS PO Mains test the Von Neumann model, the parts of the CPU, and the three system buses. Expect 1–2 questions ("Which is the brain of the computer?", "ALU does what?", "MAR holds ?"). All recall — bank them.
The Von Neumann (stored-program) model
Data and instructions share the same memory; the CPU fetches, decodes and executes in a cycle. Three units: CPU + Memory + Input/Output, linked by buses.
CPU — the "brain"
The Central Processing Unit has three parts:
- ALU (Arithmetic Logic Unit): performs arithmetic (+ − × ÷) and logic/comparison (AND, OR, NOT, >, <, =).
- CU (Control Unit): directs and coordinates all operations; runs the fetch–decode–execute cycle; does not process data itself.
- Registers: tiny, fastest storage inside the CPU.
Key registers (frequently asked)
| Register | Full form | Holds |
|---|---|---|
| PC | Program Counter | Address of the NEXT instruction |
| MAR | Memory Address Register | Address of data to be accessed |
| MDR/MBR | Memory Data/Buffer Register | The actual data/word in transit |
| IR | Instruction Register | The current instruction being executed |
| AC | Accumulator | Intermediate arithmetic/logic results |
System buses (the highways)
- Data bus: carries actual data (bi-directional).
- Address bus: carries memory addresses (uni-directional).
- Control bus: carries control/timing signals.
Other must-knows
- Motherboard: main circuit board connecting all components.
- Clock speed: measured in GHz (1 GHz = 10⁹ cycles/sec) — higher = faster.
- Microprocessor = CPU on a single chip (e.g. Intel, AMD).
Exam Tricks & Tips
- 🎯 CPU = ALU + CU + Registers.
- 🎯 Control Unit coordinates; ALU calculates. "Performs calculations" → ALU; "directs operations" → CU.
- 🎯 PC = next instruction, IR = current instruction — the classic trap pair.
- 🎯 Clock speed unit is GHz/MHz, NOT bytes.
- 🎯 Data bus is bi-directional; address bus is one-way.
- ❌ Common mistake: thinking the CU does arithmetic — it does NOT; only the ALU calculates.
Expected exam pattern
Direct MCQs on full forms (ALU, CU, MAR, PC), "brain of computer" (CPU), and bus functions. Occasionally a fetch-decode-execute ordering question.
Quick recap
CPU = ALU + CU + Registers. PC → next instruction, IR → current. Three buses: data (2-way), address (1-way), control. Von Neumann = shared memory for data + instructions.
Computer Architecture & Components — Flashcards (IBPS PO)
Cover the answer, recall, then check. 12 cards on CPU parts, registers and buses.
Q1. What are the three main parts of the CPU?
A1. ALU (Arithmetic Logic Unit), CU (Control Unit) and Registers.
Q2. What does the ALU do?
A2. Performs all arithmetic (+ − × ÷) and logical/comparison (AND, OR, NOT, >, <, =) operations.
Q3. What is the job of the Control Unit (CU)?
A3. Directs and coordinates all CPU operations via the fetch–decode–execute cycle. It does NOT perform calculations.
Q4. What does the Program Counter (PC) hold?
A4. The memory address of the NEXT instruction to be executed.
Q5. What does the Instruction Register (IR) hold?
A5. The CURRENT instruction being executed/decoded.
Q6. Which register holds the address of the memory location to be accessed?
A6. MAR — Memory Address Register.
Q7. Name the three system buses and their direction.
A7. Data bus (bi-directional), Address bus (uni-directional), Control bus (control/timing signals).
Q8. What is the "stored-program concept" / Von Neumann architecture?
A8. Data and instructions are stored together in the same memory; CPU fetches and executes them sequentially.
Q9. In what unit is CPU clock speed measured?
A9. Hertz — commonly GHz (1 GHz = 10⁹ cycles/second) or MHz.
Q10. What is a microprocessor?
A10. A CPU fabricated on a single integrated-circuit chip (e.g. Intel, AMD).
Q11. What is the accumulator (AC)?
A11. A register that stores intermediate results of ALU arithmetic/logic operations.
Q12. What is the motherboard?
A12. The main circuit board that connects the CPU, memory, and all other components of the computer.