GS Paper III · Science & Technology · ICT · Digital India
🖥 Supercomputers — The Engines of Modern Science & India's NSM
Definition · FLOPS Scale · Features · History · World Rankings · Applications · India's PARAM Series · AIRAWAT · PARAM Rudra 2024 · National Supercomputing Mission · Advantages & Disadvantages · PYQs & MCQs
🖥
What is a Supercomputer? — Definition, Features & FLOPS Scale
Definition · HPC · FLOPS · Vector Arithmetic · Multi-User · Parallel Processing
📖 Definition
A Supercomputer is a high-performance computing system capable of performing trillions of calculations per second — far beyond the capability of regular computers. They are the physical embodiment of High-Performance Computing (HPC). Performance is measured in FLOPS (Floating Point Operations Per Second), typically in petaflops (10¹⁵ FLOPS) today. They incorporate multiple CPUs working in parallel, enormous memory, and rapid data retrieval systems. Memory is typically ~250,000 times that of a normal computer. Housed in large clean rooms with high airflow for cooling.
🧠 Simple Analogy — Why Supercomputers Are Needed
A regular computer solving climate change is like one person trying to count every grain of sand on every beach in the world — would take millions of years. A supercomputer is like deploying billions of workers simultaneously, each counting different beaches at the same time. What would take a laptop 10,000 years, a modern supercomputer solves in hours. This parallel processing power is why only supercomputers can tackle problems like predicting monsoons 15 days ahead, simulating nuclear reactions, or sequencing a genome.
IBM Blue Gene/P Supercomputer — rows of computing racks in a large clean-room facility. This illustrates the physical scale of supercomputers: they typically occupy over 1,000 sq ft of floor space, consume 4+ megawatts of electricity, and require constant cooling (high air flow clean rooms). IBM developed the Roadrunner (which first broke the 1 petaFLOP barrier) and Summit supercomputers. India's NSM is building similar infrastructure indigenously. (Source: Wikimedia Commons)
PARAM Yuva II — India's supercomputer at C-DAC. India's PARAM series represents the nation's flagship supercomputing achievement. PARAM means "supreme" in Sanskrit, also stands for "PARAllel Machine." Father of Indian Supercomputer: Dr. Vijay Bhatkar (built PARAM 8000 in 1991). India now ranks among the top nations globally in HPC infrastructure. (Source: Wikimedia Commons)
⚡ Key Features of Supercomputers
🔢
Multiple CPUs (Parallel Processing)
Thousands of CPUs (and GPUs) work simultaneously. Each CPU interprets program instructions and executes arithmetic/logic operations. Break one huge problem into thousands of smaller parts → solve all simultaneously → results combined. This "parallel processing" is the fundamental difference from regular computers.
📐
Vector Arithmetic
Regular computers: perform one arithmetic operation on one pair of numbers at a time (scalar arithmetic). Supercomputers: use vector arithmetic — operate on entire lists of numbers simultaneously. Example: instead of adding A+B, B+C, C+D one at a time, a vector instruction adds [A,B,C] + [B,C,D] all at once. This massively speeds up physics simulations, weather modelling, and AI computations.
💾
Massive Storage & Memory
Memory ~250,000× a normal computer. Huge storage capacity. Crucially, rapid retrieval of stored data is essential — the CPUs must not wait for data. This is why supercomputers use specialised memory (DRAM, HBM) and high-speed interconnects (InfiniBand, Trinetra in India's indigenous system). Without fast memory access, CPUs idle → efficiency drops.
👥
Multi-User Access
Unlike personal computers, supercomputers allow multiple users to access computing resources simultaneously. Over 10,000 researchers across India access NSM supercomputers — including 1,700 PhD scholars. Jobs are queued and run according to priority — each user submits a "compute job" and the system allocates nodes accordingly.
⚡
Speed & FLOPS
Performance measured in FLOPS (Floating-Point Operations Per Second). Modern supercomputers: petaFLOPS (10¹⁵) to exaFLOPS (10¹⁸). India's AIRAWAT: 13.17 petaFLOPS peak. World fastest (Frontier, USA): 1.194 exaFLOPS — about 90× more powerful than India's best. Regular laptop: ~1 TFLOPS (teraFLOPS, 10¹²) — 13,000× slower than India's AIRAWAT.
🌡
Heat Management & Power
A supercomputer needs ~4 MW of electricity and generates enormous heat. Must be housed in large clean rooms with high air flow. Special cooling systems (liquid cooling loops, precision air conditioning) are required. Size: typically >1,000 sq ft of floor space. This is why NSM's distributed approach (multiple smaller systems across 34+ institutions) is practical for India.
📊 The FLOPS Scale — From ENIAC to Exaflop
HFLOPS
10² = 100
ENIAC (1940s) — first electronic computer
KFLOPS
10³ = 1,000
IBM 704 (1950s)
MFLOPS
10⁶ = million
CDC 6600 (1960s) — world's first supercomputer
GFLOPS
10⁹ = billion
Cray-2 (1985); PARAM 8000 India 1991 (1 GFLOPS)
TFLOPS
10¹² = trillion
ASCI Red (1990s); modern gaming GPU ~20 TFLOPS
PFLOPS
10¹⁵ = quadrillion
India AIRAWAT: 13.17 PF. Frontier: 1,194 PF
EFLOPS
10¹⁸ = quintillion
Frontier USA: 1.194 EFLOPS. The Exascale era (2022+)
🌍
World's Fastest Supercomputers — Global Rankings High Yield
Frontier USA · Fugaku Japan · Aurora · TOP500 List · Exascale Era
| Rank | Name | Country | Speed | Developer / Location | Key Purpose |
|---|---|---|---|---|---|
| #1 | Frontier | 🇺🇸 USA | 1.194 ExaFLOPS | Oak Ridge National Laboratory (ORNL), Tennessee. AMD EPYC CPUs + AMD Instinct GPUs | First exascale computer (2022). Climate modeling, materials science, genomics, nuclear simulation |
| #2 | Aurora | 🇺🇸 USA | ~1.012 ExaFLOPS | Argonne National Laboratory. Intel Xeon + Intel Ponte Vecchio GPUs | Exascale AI, cancer research, brain mapping, particle physics |
| #3 | Eagle | 🇺🇸 USA | 561 PetaFLOPS | Microsoft Azure cloud supercomputer | AI training and cloud HPC workloads |
| #4 | Fugaku | 🇯🇵 Japan | 442 PetaFLOPS | RIKEN + Fujitsu, Kobe. ARM-based A64FX processors | Drug discovery, COVID-19 research, disaster prediction, climate simulation |
| #5 | LUMI | 🇫🇮 Finland (EU) | 379 PetaFLOPS | CSC Finland (EuroHPC JU). AMD EPYC + AMD Instinct GPUs | European research — climate, AI, materials science |
| #6 | Leonardo | 🇮🇹 Italy (EU) | 239 PetaFLOPS | CINECA, Bologna. EuroHPC JU | Scientific research, drug discovery, materials |
| ~#7–8 | Summit / Sierra | 🇺🇸 USA | 200 / 125 PetaFLOPS | IBM. Oak Ridge / Lawrence Livermore NL | Nuclear simulation, climate, fusion energy |
| ~#10–15 | Sunway TaihuLight / Tianhe-2 | 🇨🇳 China | 93 / 33.86 PetaFLOPS | National Supercomputer Centers, China | AI, physics simulations, smog prediction, space research |
| #75 | AIRAWAT | 🇮🇳 India | 13.17 PetaFLOPS (peak) | C-DAC, Pune. National Supercomputing Mission | AI research and compute platform. India's fastest. Ranked 75th (June 2023) |
📊 India's Gap — and the Exascale Context
The world's fastest supercomputer (Frontier, USA) achieves 1.194 ExaFLOPS = 1,194 PetaFLOPS. India's fastest (AIRAWAT) achieves 13.17 PetaFLOPS — meaning Frontier is approximately 90× more powerful than India's best. However, context matters: India's National Supercomputing Mission (NSM) has deployed 34 supercomputers with combined capacity of 35 PetaFLOPS by March 2025 — the total national HPC capacity is growing rapidly. India's gap is mainly in semiconductor manufacturing (80%+ import dependence on advanced chips). Atmanirbhar Bharat's semiconductor mission aims to address this. For UPSC: knowing India is in the global top 500 is important; knowing the qualitative gap and its causes is equally important.
🇮🇳
India's Supercomputing Journey — PARAM Series to PARAM Rudra 2024
PARAM 8000 · Dr. Vijay Bhatkar · AIRAWAT · PARAM Rudra · Arka · Arunika
🎯 Key Reason India Built Its Own Supercomputers
1987: USA denied India's request to purchase the Cray X-MP supercomputer for academic and weather forecasting purposes — citing concerns that India might use it for nuclear weapons development. This denial became the catalyst for India's indigenous supercomputing programme. C-DAC was specifically created to develop India's own supercomputers, leading to PARAM 8000 in 1991 — a direct defiance of technology denial that became India's first supercomputing success. Father of Indian Supercomputer: Dr. Vijay Bhatkar (C-DAC). Today, India's PARAM Rudra series uses fully indigenous hardware — the "Rudra" servers are the first Made-in-India servers meeting international HPC standards.
1987 — Trigger
USA denies India's request to buy Cray X-MP supercomputer → India decides to build its own. C-DAC (Centre for Development of Advanced Computing) established.
1991 — PARAM 8000 🏆
India's first supercomputer. Built by C-DAC, Pune. Developed by Dr. Vijay Bhatkar. Gigaflops range parallel computer (1 GFLOPS). India offered to sell it to USA — a symbolic reversal. Also exhibited at the United Nations headquarters.
1993 — PARAM 9000
Peak computing power of 5 GFLOPS. India continues building indigenous HPC capability. BARC (Bhabha Atomic Research Centre) develops Anupam series. ANURAG develops PACE series.
1998 — PARAM 10000
Sustained performance of 38 GFLOPS on LINPACK benchmark. India's indigenous supercomputing programme matures.
2015 — NSM Launched
National Supercomputing Mission launched by MeitY + DST. Target: 73+ supercomputers installed across India by 2022. Budget: ₹4,500 crore ($730 million). Implemented by C-DAC (Pune) + IISc (Bengaluru).
2020 — PARAM Shivay
First supercomputer under NSM installed at IIT BHU (Varanasi). PARAM Pravega (IISc Bangalore), PARAM Utkarsh (C-DAC Bangalore), PARAM Ananta (IIT Gandhinagar), PARAM Himalaya (IIT Mandi) all deployed.
2021 — PARAM Siddhi-AI
5.267 PetaFLOPS. C-DAC, Pune. India's first AI-focused supercomputer. Jointly built by DST and MeitY under NSM. Made India's first AI-optimised HPC system.
2023 — AIRAWAT
India's fastest supercomputer. 13.17 PetaFLOPS (peak), 8.5 PetaFLOPS (Rmax). C-DAC, Pune. Ranked 75th globally in TOP500 list (June 2023). AI Research Analytics and Knowledge Dissemination platform — India's national AI compute platform.
September 2024 — PARAM Rudra Current Affairs
PM Modi virtually launched three PARAM Rudra supercomputers worth ₹130 crore under NSM. Fully indigenous — built using indigenously designed "Rudra" servers (first Made-in-India HPC servers meeting international standards). Also launched Arka and Arunika HPC systems for weather/climate research.
🆕 PARAM Rudra 2024 — India's Most Indigenous Supercomputer UPSC Current Affairs
Three deployments (September 2024):
• GMRT (Giant Metre Radio Telescope), Pune: Investigates astronomical phenomena — Fast Radio Bursts (FRBs), pulsars
• IUAC (Inter-University Accelerator Centre), Delhi: Material science and atomic physics research
• S.N. Bose Centre, Kolkata: Physics, cosmology, and earth sciences
Also launched simultaneously:
• Arka + Arunika — HPC systems at IITM (Pune) and NCMRWF (Noida) for weather/climate research → enhanced accuracy for cyclone, heavy rain, thunderstorm forecasting
• GMRT (Giant Metre Radio Telescope), Pune: Investigates astronomical phenomena — Fast Radio Bursts (FRBs), pulsars
• IUAC (Inter-University Accelerator Centre), Delhi: Material science and atomic physics research
• S.N. Bose Centre, Kolkata: Physics, cosmology, and earth sciences
Also launched simultaneously:
• Arka + Arunika — HPC systems at IITM (Pune) and NCMRWF (Noida) for weather/climate research → enhanced accuracy for cyclone, heavy rain, thunderstorm forecasting
Why PARAM Rudra is Significant:
✅ First fully indigenous supercomputer — "Rudra" servers are India's first Made-in-India HPC servers meeting international standards
✅ Indigenous system software stack (not dependent on US/EU software)
✅ Supports Atmanirbhar Bharat and Make in India in advanced technology
✅ Trinetra: India's indigenous high-speed interconnect network (100–200 Gbps) developed alongside PARAM Rudra
✅ First fully indigenous supercomputer — "Rudra" servers are India's first Made-in-India HPC servers meeting international standards
✅ Indigenous system software stack (not dependent on US/EU software)
✅ Supports Atmanirbhar Bharat and Make in India in advanced technology
✅ Trinetra: India's indigenous high-speed interconnect network (100–200 Gbps) developed alongside PARAM Rudra
NSM Status by March 2025:
• 34 supercomputers deployed
• Combined compute: 35 PetaFLOPS
• Institutions: IITs, IISc, C-DAC, R&D labs, Tier-II/III cities
• 10,000+ researchers (incl. 1,700 PhD scholars) benefiting
• 1,500+ research papers published using NSM systems
• 1 crore+ compute jobs completed
• NSM extended till December 31, 2025
• 34 supercomputers deployed
• Combined compute: 35 PetaFLOPS
• Institutions: IITs, IISc, C-DAC, R&D labs, Tier-II/III cities
• 10,000+ researchers (incl. 1,700 PhD scholars) benefiting
• 1,500+ research papers published using NSM systems
• 1 crore+ compute jobs completed
• NSM extended till December 31, 2025
| Supercomputer | Location | Speed | Focus |
|---|---|---|---|
| AIRAWAT (India's fastest) | C-DAC, Pune | 13.17 PF (peak); Rank 75 globally (TOP500, June 2023) | AI research and knowledge platform (AIRAWAT = AI Research Analytics and Knowledge) |
| PARAM Siddhi-AI | C-DAC, Pune | 5.267 PF; Rank 131 globally | AI + science; jointly built by DST + MeitY under NSM |
| Pratyush | IITM, Pune + NCMRWF, Noida | 4.006 PF; Rank 169 globally | Weather forecasting, climate research |
| Mihir | NCMRWF, Noida | 2.8 PF (2.57 Rmax); Rank 316 | Medium-range weather forecasting; Ministry of Earth Sciences |
| PARAM Rudra (3 units) | GMRT Pune, IUAC Delhi, S.N. Bose Kolkata | ~1 PF each (₹130 cr total) | Astronomy (FRBs), material science, physics, cosmology |
| Arka + Arunika | IITM Pune + NCMRWF Noida | HPC systems | Weather research — cyclone, heavy rain, thunderstorm prediction |
| PARAM Shivay | IIT BHU, Varanasi | NSM supercomputer | First system under NSM (2020). Academic research. |
| PARAM Pravega | IISc, Bengaluru | NSM — largest academic SC in India | Academic research; largest academic supercomputer in India |
| Anupam series | BARC, Mumbai | Nuclear research grade | Nuclear simulations, atomic energy research |
| PACE series | ANURAG, Hyderabad | Defence grade | Defence computing, aerodynamics, ballistic simulations |
🚀
National Supercomputing Mission (NSM) — India's HPC Flagship
MeitY + DST · C-DAC + IISc · NKN · ₹4,500 crore · Indigenous · 35 PF deployed
📋 NSM — At a Glance
Launched: 2015, by MeitY (Ministry of Electronics and IT)
Jointly led by: DST (Department of Science & Technology) + MeitY
Implemented by: C-DAC (Pune) + IISc (Bengaluru)
Budget: ₹4,500 crore ($730 million) — 7-year programme
Target: 73+ high-performance computing (HPC) facilities
Status (March 2025): 34 supercomputers deployed, 35 PetaFLOPS total
Network: National Knowledge Network (NKN) — high-speed network connecting all HPC centres and academic institutions
Extended till: December 31, 2025
Jointly led by: DST (Department of Science & Technology) + MeitY
Implemented by: C-DAC (Pune) + IISc (Bengaluru)
Budget: ₹4,500 crore ($730 million) — 7-year programme
Target: 73+ high-performance computing (HPC) facilities
Status (March 2025): 34 supercomputers deployed, 35 PetaFLOPS total
Network: National Knowledge Network (NKN) — high-speed network connecting all HPC centres and academic institutions
Extended till: December 31, 2025
🎯 NSM Objectives
- Establish a cluster of geographically distributed HPC centres connected by the National Knowledge Network (NKN)
- Enhance HPC capacity and capability in academic and R&D institutions
- Develop indigenous hardware and software for supercomputing systems (processors, networks, storage) — reducing import dependence
- Deploy three petascale supercomputers as anchor systems
- Develop HPC-aware human resources — train scientists, researchers, engineers
- Support AI, quantum computing, and big data applications for Digital India
- Enable industry access — startups and MSMEs using HPC for innovation
🏆 NSM Achievements (as of 2025)
🖥
34 Supercomputers Deployed
Combined capacity: 35 PetaFLOPS. Spread across IITs, IISc, C-DAC, R&D labs, and Tier-II/III city institutions. 85–95% system utilisation. Over 1 crore compute jobs completed.
👨🔬
10,000+ Researchers Benefited
Including 1,700 PhD scholars. 1,500+ research papers published using NSM computing resources. Research fields: drug discovery, disaster management, climate modelling, energy, astronomy, fluid dynamics, material sciences.
🏭
Indigenous Technology
PARAM Rudra: India's first fully indigenous HPC servers ("Rudra"). Trinetra: India's indigenous high-speed interconnect (100–200 Gbps). Indigenous system software stack. Supports Make in India and Atmanirbhar Bharat in advanced technology.
🏢
Industry & MSME Access
Startups and MSMEs accessing NSM HPC systems for innovation in applied research — enabling small companies to use world-class computing without owning it. Supports India's startup ecosystem in AI, biotech, materials, and climate tech.
🌦
Weather & Climate Research
Pratyush (IITM Pune + NCMRWF Noida): India's weather forecasting supercomputer system. Mihir (NCMRWF, Noida): 2.8 PF for medium-range forecasts. Arka + Arunika (2024): enhanced cyclone/storm/rain forecasting accuracy.
⚠
Challenges Remaining
India's best (AIRAWAT: 13.17 PF) vs World's best (Frontier: 1,194 PF) — gap of ~90×. Semiconductor manufacturing dependence (80%+ imported advanced chips). Need for exascale computing vision. Skilled HPC workforce gap. Commercial viability for industry.
⚙
Applications of Supercomputers — Across Critical Fields
Weather · Genome · Nuclear · Oil & Gas · Defence · Climate · AI
🌦
Weather Forecasting & Climate Research
Supercomputers process data from satellites, radars, and weather balloons → generate weather models. India: Pratyush and Mihir for Indian Ocean weather. Arka/Arunika for cyclone/storm prediction. Derecho (USA): solar geoengineering research. Frontier: climate simulation. Critical for India's disaster preparedness (cyclone warnings in Bay of Bengal).
🧬
Genome Sequencing & Drug Discovery
DNA sequencing in hours (vs days by conventional). Protein folding simulation → leads to new drug discovery. Fugaku (Japan) simulated COVID-19 drug interactions at atomic level. Stanford used supercomputers to set Guinness World Record for fastest genome sequencing. India's NSM: drug discovery as explicit application area.
🌌
Space Exploration
Simulate outer space conditions from Earth. NASA's Aitken (NASA Ames): high-resolution simulations for Artemis moon missions. Astrophysicists use supercomputers as "time machines" to simulate universe evolution. India: Pratyush used for monsoon prediction; GMRT Pune PARAM Rudra for Fast Radio Burst (FRB) astronomy research.
⚛
Nuclear Fusion & Fission Research
Simulate plasma behaviour in tokamak reactors (fusion). Virtual testing of nuclear explosions (avoiding real tests for treaty compliance). Frontier + Summit: used by General Atomics (USA) for optimising plasma performance in fusion reactors. India's BARC (Anupam series) + ANURAG (PACE series) for nuclear and defence simulations.
✈
Aviation Engineering
Detect solar flares, predict turbulence, simulate aeroelasticity (wing bending under aerodynamic forces) → design better aircraft. Frontier: used by GE Aerospace to test open fan engine architecture for next-generation commercial aircraft that reduces CO₂ emissions. India: aerospace aerodynamics research for defence aircraft design.
🛡
Military & Defence
Virtual testing of nuclear explosions (without real detonations — enables compliance with Comprehensive Test Ban Treaty). Ballistic weapons simulation. Logistics optimisation. Intelligence data analysis. India: ANURAG's PACE series dedicated to defence computing. Cyberwarfare simulation and countermeasures.
🛢
Oil, Gas & Energy Security
Process geophysical seismic data to find oil and gas reserves. Simulate subsurface geological structures. Plan drilling operations. Optimise energy extraction. For India: critical for energy security given dependence on imported oil. NSM explicitly lists energy security as a key application area.
🤖
Artificial Intelligence & Big Data
Training large AI models (LLMs, vision models) requires exaflops of compute. AIRAWAT (India's AI supercomputer): common compute platform for AI research. IndiaAI Mission (₹10,372 crore) targets AI compute through AIRAWAT. Big data mining for Digital India. NSM: startups and MSMEs using HPC for AI innovation.
🌫
Smog & Pollution Prediction
Predict fog, smog, and pollution levels in specific regions. China's Tianhe-1A used for smog prediction over Beijing. India: Delhi's severe air quality crises (AQI 999+) could benefit from supercomputer-based pollution forecasting models. India uses supercomputing for monsoon forecasting (India's economy is monsoon-dependent).
📜
PYQs & Practice MCQs
UPSC Pattern · NSM · PARAM Rudra · AIRAWAT · FLOPS · India Supercomputing
📜 UPSC Prelims Pattern — NSM Statement Type
Pattern Q
Q. Consider the following statements regarding the National Supercomputing Mission (NSM) of India:
- NSM was launched in 2015 and is jointly led by the Ministry of Electronics and IT (MeitY) and the Department of Science and Technology (DST).
- The mission is implemented by C-DAC (Pune) and IISc (Bengaluru) and aims to build a network of supercomputers connected by the National Knowledge Network (NKN).
- PARAM Shivay, the first supercomputer under NSM, was installed at IIT BHU, Varanasi.
- PARAM Rudra, launched in September 2024, uses entirely imported hardware and software to maintain international quality standards.
- a) 1, 2 and 3 only
- a) 1, 2 and 3 only ✓
- b) 2 and 4 only
- c) 1, 3 and 4 only
- d) 1, 2, 3 and 4
Statement 1 CORRECT: NSM launched 2015. Jointly led by MeitY + DST. Budget ₹4,500 crore.
Statement 2 CORRECT: C-DAC Pune + IISc Bengaluru implement NSM. Systems connected via National Knowledge Network (NKN) for data sharing across institutions. As of March 2025: 34 supercomputers, 35 PetaFLOPS deployed.
Statement 3 CORRECT: PARAM Shivay was the first supercomputer installed under NSM (2020) at IIT BHU, Varanasi. Subsequent installations include PARAM Pravega (IISc Bangalore), PARAM Utkarsh (C-DAC Bangalore), PARAM Ananta (IIT Gandhinagar), PARAM Himalaya (IIT Mandi).
Statement 4 WRONG — Key Trap! PARAM Rudra (September 2024) is India's most INDIGENOUS supercomputer. It is built using indigenously designed and manufactured "Rudra" servers — the first Made-in-India HPC servers meeting international standards — along with an indigenously developed system software stack. The whole point of PARAM Rudra was to demonstrate self-reliance (Atmanirbhar Bharat) in advanced computing hardware. Using imported hardware would contradict the core objective.
Statement 2 CORRECT: C-DAC Pune + IISc Bengaluru implement NSM. Systems connected via National Knowledge Network (NKN) for data sharing across institutions. As of March 2025: 34 supercomputers, 35 PetaFLOPS deployed.
Statement 3 CORRECT: PARAM Shivay was the first supercomputer installed under NSM (2020) at IIT BHU, Varanasi. Subsequent installations include PARAM Pravega (IISc Bangalore), PARAM Utkarsh (C-DAC Bangalore), PARAM Ananta (IIT Gandhinagar), PARAM Himalaya (IIT Mandi).
Statement 4 WRONG — Key Trap! PARAM Rudra (September 2024) is India's most INDIGENOUS supercomputer. It is built using indigenously designed and manufactured "Rudra" servers — the first Made-in-India HPC servers meeting international standards — along with an indigenously developed system software stack. The whole point of PARAM Rudra was to demonstrate self-reliance (Atmanirbhar Bharat) in advanced computing hardware. Using imported hardware would contradict the core objective.
🧪 Practice MCQs — Supercomputers (Click to attempt)
Q1. The performance of supercomputers is measured in FLOPS (Floating-Point Operations Per Second). Which of the following correctly arranges the FLOPS units in ascending order of magnitude?
- (a) GFLOPS → TFLOPS → PFLOPS → EFLOPS (Giga → Tera → Peta → Exa)
- (b) TFLOPS → GFLOPS → PFLOPS → EFLOPS
- (c) PFLOPS → GFLOPS → TFLOPS → EFLOPS
- (d) GFLOPS → PFLOPS → TFLOPS → EFLOPS
The FLOPS scale in ascending order: HFLOPS (10²) → KFLOPS (10³) → MFLOPS (10⁶) → GFLOPS (10⁹, Giga) → TFLOPS (10¹², Tera) → PFLOPS (10¹⁵, Peta) → EFLOPS (10¹⁸, Exa). India's AIRAWAT peak: 13.17 PFLOPS. World's fastest (Frontier, USA): 1.194 EFLOPS = 1,194 PFLOPS. This means Frontier is about 90× more powerful than India's best. Context: India's PARAM 8000 (1991): 1 GFLOPS. Modern laptop: ~1–10 TFLOPS. AIRAWAT: 13.17 PFLOPS. Frontier: 1,194 PFLOPS. The progression shows how supercomputing performance has grown by a factor of ~10¹⁶ (ten thousand trillion times) from ENIAC to Frontier — 80 years of exponential growth.
Q2. The development of India's first supercomputer PARAM 8000 in 1991 was directly triggered by:
- (a) India's decision to participate in the Cold War arms race by developing independent nuclear simulation capability
- (b) The USA's refusal in 1987 to sell India the Cray X-MP supercomputer for academic and weather forecasting purposes, compelling India to develop its own indigenous supercomputing programme under C-DAC
- (c) The success of India's 1983 SLV-3 rocket launch which demonstrated India's technological capability and motivated C-DAC to enter the supercomputing domain
- (d) A recommendation by the National Planning Commission's Seventh Five Year Plan (1985-90) that India should build five supercomputers by 1990
In 1987, the USA denied India's request to purchase the Cray X-MP supercomputer. The official reason given was concern that India might use the supercomputer for nuclear weapons research (given India's nuclear programme). The USA applied export controls restricting high-performance computing transfers to India under the Cold War-era COCOM (Coordinating Committee for Multilateral Export Controls) framework. This technology denial had the paradoxical effect of galvanising India's indigenous technology development. C-DAC (Centre for Development of Advanced Computing) was established specifically to develop India's own supercomputer. Under Dr. Vijay Bhatkar ("Father of the Indian Supercomputer"), C-DAC developed PARAM 8000 in 1991 — achieving gigaflops-range performance. India then turned the tables by offering to sell PARAM 8000 to the USA and exhibiting it at the United Nations — a symbolic statement of technological capability. PARAM 8000 was also installed in various developing countries. This story illustrates how technology denial can become a catalyst for indigenous innovation — a lesson relevant to India's current efforts in semiconductors, quantum computing, and 6G.
Q3. Three PARAM Rudra supercomputers were launched in September 2024. Which of the following correctly matches each PARAM Rudra unit with its deployment location and research focus?
- (a) Pune (IITM) — weather forecasting; Delhi (IIT Delhi) — materials science; Kolkata (IIT Kharagpur) — aerospace
- (b) Mumbai (BARC) — nuclear research; Delhi (DRDO) — defence; Bengaluru (IISc) — biological sciences
- (c) Pune (GMRT — Giant Metre Radio Telescope) — astronomy/Fast Radio Bursts; Delhi (IUAC — Inter-University Accelerator Centre) — material science and atomic physics; Kolkata (S.N. Bose Centre) — physics, cosmology, earth sciences
- (d) Pune (C-DAC) — AI research; Delhi (NIC) — e-governance; Kolkata (ISRO) — satellite data processing
The three PARAM Rudra units (September 2024, PM Modi inauguration, ₹130 crore total) are: (1) GMRT (Giant Metre Radio Telescope) in Pune — will investigate astronomical phenomena including Fast Radio Bursts (FRBs), pulsars, and other radio astronomy research. FRBs are brief, intense bursts of electromagnetic radiation in radio frequencies from deep space — one of astronomy's most mysterious phenomena. (2) IUAC (Inter-University Accelerator Centre) in New Delhi — enhances research in material science, atomic physics, accelerator-based research, and nuclear structure studies. (3) S.N. Bose National Centre for Basic Sciences in Kolkata — drives advanced research in physics, cosmology, quantum physics, and earth sciences. Named after Satyendra Nath Bose (of the Bose-Einstein statistics). Additionally, simultaneously launched: Arka (at IITM Pune) and Arunika (at NCMRWF Noida) — dedicated HPC systems for weather and climate research, enhancing India's tropical cyclone, heavy rain, and thunderstorm forecasting capability. The Rudra servers used in PARAM Rudra are India's first indigenously designed and manufactured HPC servers meeting international standards — a significant Atmanirbhar Bharat achievement.
Q4. "Vector Arithmetic" in supercomputers is best described as:
- (a) The use of graphical processing units (GPUs) to render 3D vector graphics for scientific visualisation — enabling researchers to see simulation results as 3D images
- (b) A method of storing data in directional vector format (pointing from source to destination) that enables supercomputers to handle network routing computations faster than regular computers
- (c) A mathematical technique borrowed from quantum mechanics that allows supercomputers to perform calculations using complex numbers with both magnitude and direction
- (d) The ability to operate on entire lists (arrays/vectors) of numbers simultaneously in a single instruction — instead of processing one pair of numbers at a time — dramatically accelerating repetitive mathematical operations like those in weather models, physics simulations, and AI training
Vector arithmetic (or vector processing) is the capability of a processor to apply a single instruction to multiple data elements simultaneously. In regular (scalar) computing: to add three pairs (A+B, C+D, E+F), a processor performs three separate operations sequentially. In vector computing: a single "vector ADD" instruction processes [A,C,E] + [B,D,F] in one step. This is related to but distinct from GPU rendering (option a), network routing (option b), or quantum mechanics (option c). Why it matters: scientific computing is full of repetitive operations on large datasets. Weather modeling: adding temperature gradients across millions of grid cells. Image processing: applying the same filter to every pixel. Physics simulation: updating positions and velocities of millions of particles. Matrix multiplication in AI: applying the same math to every element of huge matrices. Vector arithmetic was first successfully implemented in the Cray-1 (1976) — the document explicitly mentions this as a key historical milestone. Modern supercomputers combine vector CPUs with massively parallel GPU accelerators (which are essentially extreme vector processors) to achieve petaflops/exaflops performance. India's PARAM series and global supercomputers like Frontier all rely on vectorised operations as a core architectural principle.
Q5. AIRAWAT, India's fastest supercomputer, is significant in the context of India's technological development because:
- (a) It is the world's fastest supercomputer, exceeding even the USA's Frontier in computational performance
- (b) It was India's first supercomputer to break the one petaFLOP barrier, marking India's entry into the petascale era of computing
- (c) It was built using only Indian components and software without any foreign technology, demonstrating complete technological self-sufficiency in high-performance computing
- (d) As India's fastest supercomputer (ranked 75th globally in the TOP500 list as of June 2023 with 13.17 peak petaFLOPS), it serves as India's national AI compute platform, providing shared HPC resources for AI research under the IndiaAI Mission and National Supercomputing Mission, while highlighting India's gap with world leaders (Frontier at ~1,194 PF) that needs to be bridged
AIRAWAT (AI Research Analytics and Knowledge Dissemination Platform) is India's fastest supercomputer as of June 2023: peak performance of 13.17 PetaFLOPS, sustained 8.5 PetaFLOPS (Rmax), globally ranked 75th on the TOP500 supercomputer list. It is located at C-DAC Pune and was jointly developed under DST + MeitY as part of the National Supercomputing Mission. AIRAWAT's significance: (1) India's national AI compute platform — provides shared computing resources for AI researchers across institutions; (2) Supports IndiaAI Mission (₹10,372 crore) which uses AIRAWAT as a key compute infrastructure component; (3) Represents India's entry into the AI supercomputing era (not just traditional HPC). However, option (d) correctly captures both the significance AND the realistic context: India's AIRAWAT (13.17 PF) is approximately 90× less powerful than the world's fastest, Frontier (1,194 PF = ~1.194 ExaFLOPS). Option (a) is wrong — AIRAWAT is not the world's fastest (Frontier, USA is). Option (b) is wrong — PARAM Siddhi-AI (5.267 PF) was India's first to clearly break one petaFLOP; Pratyush (4 PF) also preceded AIRAWAT. Option (c) is wrong — while NSM emphasises indigenous development, AIRAWAT uses NVIDIA GPUs and other international components; it is PARAM Rudra (2024) that uses the most indigenous hardware (Rudra servers).
⚡ Quick Revision — Supercomputers Summary
| Topic | Key Facts to Remember |
|---|---|
| Definition | High-performance computer with trillions of calculations/second. Performance: FLOPS (Floating-Point Operations Per Second). Memory ~250,000× normal computer. Housed in clean rooms with high airflow cooling. ~4 MW electricity consumption. >1,000 sq ft space. |
| FLOPS Scale | GFLOPS (10⁹, 1980s Cray-2) → TFLOPS (10¹², 1990s ASCI Red) → PFLOPS (10¹⁵, 2010s Jaguar) → EFLOPS (10¹⁸, 2020s Frontier). India AIRAWAT: 13.17 PFLOPS. World fastest Frontier: 1.194 EFLOPS = 1,194 PFLOPS. |
| Key Features | Multiple CPUs (parallel processing) · Vector arithmetic (operates on lists of numbers, not just pairs) · Massive storage and memory · Multi-user access · High computation speed |
| World Rankings | #1 Frontier (USA): 1.194 ExaFLOPS (ORNL, AMD) · #2 Aurora (USA): 1.012 EF · #4 Fugaku (Japan): 442 PF (RIKEN+Fujitsu) · #5 LUMI (Finland/EU): 379 PF · Summit/Sierra (USA): 200/125 PF · Sunway TaihuLight (China): 93 PF · India AIRAWAT: 13.17 PF (Rank 75) |
| India History | 1987: USA denied Cray X-MP → India builds own. 1991: PARAM 8000 (1 GFLOPS, Dr. Vijay Bhatkar, C-DAC) — India's first. 1993: PARAM 9000. 1998: PARAM 10000. 2015: NSM launched. 2020: PARAM Shivay (IIT BHU — first under NSM). 2021: PARAM Siddhi-AI (5.267 PF). 2023: AIRAWAT (13.17 PF, Rank 75). 2024: PARAM Rudra (3 units, ₹130 cr, fully indigenous). |
| PARAM Rudra 2024 | PM Modi launched Sept 2024. 3 units: GMRT Pune (astronomy/FRBs), IUAC Delhi (material science/atomic physics), S.N. Bose Kolkata (physics/cosmology). Fully indigenous Rudra servers (Made-in-India HPC servers). Trinetra: indigenous interconnect (100–200 Gbps). Also: Arka + Arunika (IITM Pune + NCMRWF Noida) for weather/climate research. |
| NSM | 2015, MeitY + DST jointly led. C-DAC + IISc implementing. ₹4,500 crore. Target: 73+ HPC centres on NKN (National Knowledge Network). March 2025: 34 supercomputers, 35 PetaFLOPS, 10,000+ researchers, 1,700 PhD scholars, 1,500+ papers, 1 crore+ compute jobs. Extended till Dec 2025. |
| Applications | Weather/climate forecasting (Pratyush, Mihir, Arka, Arunika) · Genome sequencing/drug discovery · Space exploration (NASA Aitken for Artemis) · Nuclear fusion/fission simulation · Aviation engineering (GE Aerospace + Frontier) · Oil/gas exploration · Military (virtual nuclear testing) · AI/Big Data (AIRAWAT) · Smog prediction (Tianhe-1A, China) |
| Disadvantages | Physical size (>1,000 sq ft) · High power consumption (~4 MW) · Special maintenance expertise needed · Enormous storage requirements · Heat generation damages components · Very expensive (ordinary users cannot buy) |
🚨 5 UPSC Traps — Supercomputers:
Trap 1 — "India's PARAM Rudra (2024) uses imported hardware to maintain international quality standards" → WRONG! PARAM Rudra is India's most indigenous supercomputer. The "Rudra" servers used in PARAM Rudra are India's first indigenously designed and manufactured HPC servers meeting international standards — the opposite of imported. The indigenous system software stack was also developed domestically. This is a deliberate Atmanirbhar Bharat achievement. Similarly, Trinetra (India's indigenous high-speed interconnect, 100–200 Gbps) was developed alongside PARAM Rudra. This question was specifically designed to test whether students understand the significance of NSM's indigenisation objective.
Trap 2 — "PARAM Shivay was the first supercomputer in India" → WRONG (misleading)! PARAM Shivay (2020) was the first supercomputer installed under the National Supercomputing Mission (NSM) — not India's first supercomputer ever. India's first supercomputer was PARAM 8000, developed by C-DAC under Dr. Vijay Bhatkar in 1991 — 29 years before PARAM Shivay. This is a classic UPSC trap: the qualifier "under NSM" is what makes PARAM Shivay significant, not being India's first.
Trap 3 — "AIRAWAT is India's latest/newest supercomputer" → WRONG! AIRAWAT (2023) is India's fastest supercomputer — not the latest (newest). PARAM Rudra (launched September 2024) and the Arka/Arunika HPC systems (also launched September 2024) are newer than AIRAWAT. AIRAWAT is the fastest in terms of performance (13.17 PetaFLOPS peak, TOP500 rank 75), but PARAM Rudra is the most recent deployment and most indigenous. Always distinguish between "fastest" and "latest."
Trap 4 — "Supercomputers measure performance in MIPS (Million Instructions Per Second)" → WRONG! Modern supercomputer performance is measured in FLOPS (Floating-Point Operations Per Second) — not MIPS. MIPS (Million Instructions Per Second) was used for early computers like the CDC 6600 (3 MIPS) in the 1960s. For scientific computing (which involves complex mathematical operations on decimal numbers), floating-point performance (FLOPS) is the relevant metric. The TOP500 supercomputer list uses the LINPACK benchmark measured in FLOPS (specifically Rmax = sustained FLOPS on the LINPACK benchmark and Rpeak = theoretical peak FLOPS).
Trap 5 — "The Cray-1 (1976) was the world's first supercomputer" → WRONG! The world's first supercomputer was the CDC 6600 (1964), designed by Seymour Cray, with a performance of 3 MIPS — the fastest machine of its time. The Cray-1 (1976) was significant for being the first successful implementation of vector processing — which is a different milestone. Seymour Cray built both: CDC 6600 (world's first supercomputer) and then Cray-1 at his own company. The 1985 Cray-2 was the first to exceed 1 billion FLOPS (1 GFLOPS). These are three distinct UPSC-relevant milestones: CDC 6600 (1964, first SC), Cray-1 (1976, first vector processor), Cray-2 (1985, first 1+ GFLOPS).
Trap 1 — "India's PARAM Rudra (2024) uses imported hardware to maintain international quality standards" → WRONG! PARAM Rudra is India's most indigenous supercomputer. The "Rudra" servers used in PARAM Rudra are India's first indigenously designed and manufactured HPC servers meeting international standards — the opposite of imported. The indigenous system software stack was also developed domestically. This is a deliberate Atmanirbhar Bharat achievement. Similarly, Trinetra (India's indigenous high-speed interconnect, 100–200 Gbps) was developed alongside PARAM Rudra. This question was specifically designed to test whether students understand the significance of NSM's indigenisation objective.
Trap 2 — "PARAM Shivay was the first supercomputer in India" → WRONG (misleading)! PARAM Shivay (2020) was the first supercomputer installed under the National Supercomputing Mission (NSM) — not India's first supercomputer ever. India's first supercomputer was PARAM 8000, developed by C-DAC under Dr. Vijay Bhatkar in 1991 — 29 years before PARAM Shivay. This is a classic UPSC trap: the qualifier "under NSM" is what makes PARAM Shivay significant, not being India's first.
Trap 3 — "AIRAWAT is India's latest/newest supercomputer" → WRONG! AIRAWAT (2023) is India's fastest supercomputer — not the latest (newest). PARAM Rudra (launched September 2024) and the Arka/Arunika HPC systems (also launched September 2024) are newer than AIRAWAT. AIRAWAT is the fastest in terms of performance (13.17 PetaFLOPS peak, TOP500 rank 75), but PARAM Rudra is the most recent deployment and most indigenous. Always distinguish between "fastest" and "latest."
Trap 4 — "Supercomputers measure performance in MIPS (Million Instructions Per Second)" → WRONG! Modern supercomputer performance is measured in FLOPS (Floating-Point Operations Per Second) — not MIPS. MIPS (Million Instructions Per Second) was used for early computers like the CDC 6600 (3 MIPS) in the 1960s. For scientific computing (which involves complex mathematical operations on decimal numbers), floating-point performance (FLOPS) is the relevant metric. The TOP500 supercomputer list uses the LINPACK benchmark measured in FLOPS (specifically Rmax = sustained FLOPS on the LINPACK benchmark and Rpeak = theoretical peak FLOPS).
Trap 5 — "The Cray-1 (1976) was the world's first supercomputer" → WRONG! The world's first supercomputer was the CDC 6600 (1964), designed by Seymour Cray, with a performance of 3 MIPS — the fastest machine of its time. The Cray-1 (1976) was significant for being the first successful implementation of vector processing — which is a different milestone. Seymour Cray built both: CDC 6600 (world's first supercomputer) and then Cray-1 at his own company. The 1985 Cray-2 was the first to exceed 1 billion FLOPS (1 GFLOPS). These are three distinct UPSC-relevant milestones: CDC 6600 (1964, first SC), Cray-1 (1976, first vector processor), Cray-2 (1985, first 1+ GFLOPS).
