Supercomputers

Unlike traditional Computers, supercomputers use more than one central processing unit (CPU).
This allows supercomputers to use parallel processing instead of serial processing.
Parallel processing means that the supercomputer splits tasks into pieces and distributes the pieces between the CPUs so that they can be worked on at once.

Note

Because of modern supercomputers' power consumption, data centers require cooling systems and suitable facilities to house it all. Supercomputers are often used to run artificial intelligence (AI) programs.

Types of Supercomputers

Supercomputer Description
Vector Supercomputers Use vector processors for high-speed data processing.
#Cluster Supercomputers Use clusters of regular computers connected by a network.
MIMD Supercomputers Process different instructions on different data simultaneously.
MPP Supercomputers Consist of thousands of processors with independent memory.
#SMP Supercomputers Processors share memory and are controlled by a single OS.
#Grid Supercomputers Distributed across multiple locations via the internet.
Hybrid Supercomputers Combine CPUs and GPUs for improved performance.
Exascale Supercomputers Next-generation systems performing at exaflop speeds.
#Quantum Supercomputers Use qubits to solve specific types of problems.
ASIC Supercomputers Custom-designed for specific computational tasks.

Performance

Supercomputing is measured in floating-point operations per second (FLOPS).
Floating-point numbers have decimal points in them, e.g., 1.0. By contrast, the number 1 (without a decimal point) is a binary integer used to perform binary integer operations (+, -, *, /).

Note

The performance of quantum supercomputers is not measured in FLOPS! Quantum computers use the principles of quantum mechanics and Qubits to process information. They are not faster because of its hardware, but because of new possibilities from quantum algorithms.

Unit FLOPS Example Decade
Hundred FLOPS 100 = 102 Eniac ~1940s
KFLOPS (kiloflops) 1.000 = 103 IBM 704 ~1950s
MFLOPS (megaflops) 1.000.000 = 106 CDC 6600 ~1960s
GFLOPS (gigaflops) 1.000.000.000 = 109 Cray-2 ~1980s
TFLOPS (teraflops) 1.000.000.000.000 = 1012 ASCI Red ~1990s
PFLOPS (petaflops) 1.000.000.000.000.000 = 1015 Summit ~2010s
EFLOPS (exaflops) 1.000.000.000.000.000.000 = 1018 Frontier ~2020s

Cluster Supercomputers

You can make a supercomputer by making multiple processors cooperate to work on a complex problem through parallel processing. Alternatively, you could use regular PCs and interconnect them using a very fast local area network (LAN) so they work in a similar way. That kind of supercomputer is called a cluster.

Example

Google web searches are processed with clusters of regular computers dotted in data centers around the world.

Grid Supercomputers

A grid is a supercomputer made up of separate computers, similarly to #Clusters. The difference is that the computers are connected through the Internet or other computer networks (distributed computing). This means that the power of a computer is spread across multiple locations instead of being located in one, single place (tcentralized computing).

Example

The CERN Worldwide LHC Computing Grid, assembled to process data from the LHC (Large Hadron Collider) particle accelerator, uses a supercomputer grid. Here, a dozen powerful mainframe computers in universities are linked together by a network to form the supercomputer grid. However, not all the computers are actively working in the grid all the time.

SMP Supercomputers

The CPUs that make up a symmetric multiprocessing (SMP) supercomputer are grouped into compute nodes, comprising a processor or a group of processors and a memory block. At scale, a supercomputer can contain tens of thousands of nodes. With interconnect communication capabilities, these nodes can collaborate on solving a specific problem. Nodes also use interconnects to communicate with I/O systems, like data storage and networking.

Quantum Supercomputers

A quantum supercomputer uses the principles of quantum mechanics (superposition, entanglement, and interference) to perform calculations in ways that are fundamentally different from regular computers. By manipulating qubits, quantum gates, and circuits, these systems can solve certain problems far more efficiently than their classical counterparts.

Note

Quantum supercomputers are still in the experimental stage for most applications.

Qubits

The basic unit of information in a quantum computer is the quantum bit or qubit. Unlike classical bits, which can be either 0 or 1, qubits can exist in a state of 0, 1, or any quantum superposition of these states (meaning a qubit can represent both 0 and 1 simultaneously). This allows quantum computers to process a vast number of possibilities at once.

Type of Qubit Description
Superconducting Circuits Use superconducting loops that carry persistent currents.
Trapped Ions Use ions trapped in an electromagnetic field, manipulated by lasers.
Topological Qubits Use exotic particles called anyons, with inherent fault tolerance.
Photonic Qubits Use particles of light (photons) to represent and manipulate quantum information.