Master10
Science & Technology Module

Supercomputing & High-Performance Computing

Supercomputing relies on massively parallel processing architectures to execute trillions of high-precision floating-point operations per second (FLOPS), measured biannually by the global TOP500 benchmark. Frontier, deployed at the Oak Ridge National Laboratory in the United States, became the world's first officially verified exascale supercomputer in 2022, exceeding one quintillion calculations per second (

Key Concepts & Examination Highlights

  • Frontier at Oak Ridge National Laboratory became the world's first verified exascale supercomputer on the TOP500 list in June 2022.
  • India's National Supercomputing Mission (NSM) is jointly implemented by C-DAC and the Indian Institute of Science (IISc).
  • AIRAWAT, installed at C-DAC Pune, is India's leading AI-dedicated high-performance supercomputing system on the TOP500 rankings.
  • Exascale computing represents systems capable of executing at least one exaFLOPS (101810^{18} floating-point operations per second), verified by the standard High-Performance LINPACK (HPL) benchmark.
  • PARAM 8000, built indigenously by the Centre for Development of Advanced Computing (C-DAC) under Dr. Vijay Bhatkar in 1991, was India's first supercomputer.
  • PARAM Pravega, deployed at the Indian Institute of Science (IISc) Bengaluru under the NSM, delivers a peak supercomputing capacity of 3.3 petaFLOPS.
  • PARAM Siddhi-AI, built on NVIDIA DGX SuperPOD architecture at C-DAC Pune, is one of India's fastest high-performance AI supercomputers.
  • Fugaku, developed jointly by RIKEN and Fujitsu in Japan, utilized ARM-based A64FX processors and held the world's number one TOP500 spot prior to Frontier.
  • Supercomputing performance is evaluated using the LINPACK benchmark for dense matrix computations and the Green500 ranking for energy efficiency (FLOPS per watt).
  • Under Phase III of the National Supercomputing Mission, India focuses on indigenous manufacturing of high-performance server boards, direct contact liquid cooling, and the Rudra server platform.
  • High-Performance Computing (HPC) architectures rely heavily on Graphics Processing Units (GPUs) and specialized Tensor Processing Units (TPUs) alongside CPUs for massive parallel workload execution.
  • Aurora, deployed at the Argonne National Laboratory in the United States, achieved an exascale performance exceeding 1.01 exaFLOPS on Intel Xeon CPU Max and Data Center GPU Max hardware.
  • LUMI, located in Kajaani, Finland, is Europe's fastest supercomputer, powered by AMD EPYC CPUs and AMD Instinct MI250X GPUs under the EuroHPC Joint Undertaking.
  • India's National Supercomputing Mission was launched in 2015 with an initial financial outlay of Rs 4,500 crore to build a nationwide HPC grid connecting academic and research institutions.
  • PARAM Ananta, commissioned at IIT Gandhinagar under the NSM, provides 838 teraFLOPS of computing capacity for computational chemistry and climate modeling.
  • PARAM Ganga, installed at IIT Roorkee, delivers an HPC capacity of 1.66 petaFLOPS using C-DAC's indigenously developed computing architecture.
  • PARAM Brahma, deployed at IISER Pune, was among the first NSM supercomputers to utilize direct liquid cooling to achieve high operational energy efficiency.
  • Pratyush (at IITM Pune) and Mihir (at NCMRWF Noida) are dedicated High-Performance Computing systems used by the Ministry of Earth Sciences for weather and monsoon forecasting.
  • The Rudra server board, developed indigenously by C-DAC, is a dual-socket x86 computing node designed to support high-density HPC clusters.
  • The Trinetra interconnect, developed by C-DAC, provides high-bandwidth, low-latency inter-node communication across large-scale distributed supercomputing clusters.
  • Heterogeneous HPC systems combine multi-core CPUs for serial control tasks with massive parallel GPU accelerators to maximize compute density per rack.
  • Quantum-classical hybrid supercomputing integrates superconducting quantum processing units (QPUs) with classical exascale clusters to accelerate complex combinatorial optimization.
  • The Top500 list is published twice annually (in June at ISC and November at SC) using the standard HPL LINPACK benchmark to rank the world's most powerful non-distributed computer systems.
  • Green500 rankings evaluate supercomputers on energy efficiency measured in gigaFLOPS per watt, promoting sustainable supercomputing with liquid immersion cooling technologies.
  • Molecular dynamics simulation software such as GROMACS, NAMD, and AMBER are extensively parallelized to run across thousands of GPU cores in supercomputing clusters.
  • Supercomputer interconnect technologies such as NVIDIA InfiniBand (Quantum-2 NDR) and HPE Slingshot provide ultra-low sub-microsecond latencies and multi-terabit bandwidth across compute clusters.
  • The Message Passing Interface (MPI) is the standardized and portable message-passing system designed for distributed parallel computing on high-performance supercomputing clusters.
  • OpenMP (Open Multi-Processing) provides a portable, scalable programming model for shared-memory parallel multiprocessing across multi-core supercomputing nodes.
  • C-DAC's PARAM Shavig is an indigenously developed cloud-based HPC framework providing supercomputing access to researchers across Indian academic institutions.
  • PARAM Utkarsh, deployed at C-DAC Bengaluru, is an HPC AI facility delivering scalable deep learning training infrastructure under the National Supercomputing Mission.
  • PARAM Seva, installed at IIT Hyderabad, provides specialized computing nodes for electronic structure calculations and materials discovery.
  • The Tianhe-2A supercomputer, developed by China's National University of Defense Technology (NUDT), utilizes custom Matrix-2000 coprocessors and TH-Express-2 interconnects.
  • Sunway TaihuLight, located in Wuxi, China, utilized 40,960 indigenously designed 260-core ShenWei SW26010 manycore processors without external accelerator cards.
  • Summit, built by IBM for Oak Ridge National Laboratory, held the world number one spot on TOP500 from 2018 to 2020 using IBM POWER9 CPUs and NVIDIA Volta V100 GPUs.
  • Direct-to-chip liquid cooling circuits circulate dielectric coolant directly across CPU and GPU cold plates, removing over 90% of heat without relying on energy-intensive air conditioning.
  • The LINPACK Benchmark measures how fast a dedicated computer system solves a dense system of linear equations (Ax=bAx = b), expressing throughput in floating-point operations per second (FLOPS).
  • The High-Performance Conjugate Gradients (HPCG) benchmark evaluates supercomputers on memory-bandwidth-constrained sparse linear equation solvers to complement LINPACK ratings.
  • The IO500 benchmark ranks high-performance computing storage architectures based on read/write bandwidth and metadata input/output operations per second (IOPS).
  • Lustre is an open-source, massively parallel distributed file system designed to handle petabyte-scale storage clusters and multi-terabyte per second I/O throughput in HPC environments.
  • IBM GPFS (General Parallel File System), known as Spectrum Scale, provides scalable high-throughput concurrent shared disk access across thousands of supercomputer nodes.
  • PARAM Shivay was the first supercomputer commissioned under the National Supercomputing Mission, installed at IIT-BHU Varanasi with 833 teraFLOPS computing power.
  • PARAM Porul is a state-of-the-art supercomputer established at NIT Tiruchirappalli under Phase II of the NSM to support environmental research and biomedical engineering.
  • PARAM Smriti, installed at NABI Mohali, is dedicated to computational genomics, agri-food biotechnology, and agricultural bioinformatics.
  • CUDA (Compute Unified Device Architecture), developed by NVIDIA in 2006, is a parallel computing platform and API that allows software developers to use GPUs for general-purpose processing (GPGPU).
  • ROCm (Radeon Open Compute) is AMD's open-source software development platform for GPU-accelerated high-performance computing and machine learning.
  • Quantum computing simulation on classical supercomputers requires exponential memory scaling, where simulating a 50-qubit quantum state demands petabytes of RAM.
  • Neuromorphic supercomputing architectures, like SpiNNaker and Intel Loihi, mimic biological neural spikes using asynchronous circuits for ultra-energy-efficient parallel computation.
  • Fault tolerance in exascale supercomputers relies on checkpoint-restart mechanisms and asynchronous multi-level checkpointing to protect long-running simulations from node failures.
  • Power usage effectiveness (PUE) measures the energy efficiency of supercomputing data centers, defined as total facility energy divided by IT equipment energy.
  • The Rudra server ecosystem developed by C-DAC integrates secure hardware root-of-trust, indigenously designed motherboards, and power management firmware.
Curriculum & Reference Sources: TOP500 Global Benchmark, Centre for Development of Advanced Computing (C-DAC), Department of Science and Technology (DST)

Sample Solved Questions & Concept Explanations

8 Verified Concept Questions
Q1.EASY

Which high-speed computing system is specifically designed to perform quadrillions of floating-point operations per second (FLOPS)?

Q2.EASY

What is India's fastest supercomputer installed at CDAC Pune under the National Supercomputing Mission (NSM)?

Q3.MEDIUM

What is the speed benchmark used to measure supercomputing performance on the TOP500 global list?

Q4.MEDIUM

The 'Frontier' supercomputer at Oak Ridge National Laboratory (ORNL) achieved what historical milestone in 2022?

Q5.MEDIUM

Which organization developed the PARAM series of Indian supercomputers, starting with PARAM 8000 in 1991?

Q6.EASY

Which programming language, created by Guido van Rossum and released in 1991, is widely celebrated for its clean readable syntax and dominant use in AI and data science?

Q7.EASY

What does the acronym "HTTP" stand for in web browsing and internet protocols?

Q8.EASY

What is an autonomous aircraft piloted without a human onboard, commonly operated via remote control or programmed software, called?