The multinational company Meta Platforms, in partnership with Nvidia, has announced a super-powerful computer based on artificial intelligence technology. It is named the AI Research SuperCluster (RSC) and has been designed for the development and implementation of neural networks.
Meta Platforms’ Latest Development – The Most Powerful AI Computer
Meta, known for its technological projects, has introduced the world’s most powerful computer based on Nvidia’s GGX A100 graphics processors. The system is called AI Research SuperCluster (RSC). This supercomputer is designed for scientific research in the field of artificial intelligence, neural network development, and machine learning.
“Meta’s developments have led to the creation of the world’s most powerful computing device capable of executing deep learning algorithms for artificial intelligence,” said Meta CEO Mark Zuckerberg. “RSC has been developed using cutting-edge technologies and algorithms that can change the world as early as 2022,” he stated.
In 2017, Meta Platforms announced a business partnership with Nvidia. Meta developers utilized GPUs with GV100 hardware architecture to assemble a high-tech system consisting of 22,000 such processors. During this time, companies like Microsoft and Nvidia launched their own supercomputers. However, Meta aimed to create something entirely new, and by the end of 2022, its main goal was achieved.
During the research, it was found that the device can process 16 terabytes of data per second, which far exceeds the processing speed of other supercomputers. This is due to the fact that the GV100 processors, which form the foundation of Meta’s computer, can operate at a frequency of 1455 MHz. The system itself functions on 6080 accelerators, with the NVIDIA Quantum-2 InfiniBand platform serving as the connecting link.
Goals of the RSC Development
According to the company, this innovation has no analogs in the world. The machine is intended to perform complex scientific and industrial tasks. The super-powerful computer is designed for AI research, including:
- Creating and training thousands of virtual machines in the cloud;
- Finding potential solutions related to algorithm errors in AI and machine learning systems;
- Utilizing open-source AI systems provided by Open Source;
- Processing data, predicting, and making decisions related to cybersecurity;
- Supporting Data Science, Data Analytics, Computer Vision, and other information activities;
- Solving problems related to image recognition, data analysis, modeling, and much more.
In particular, it will be used in areas such as machine vision, robotics, and other fields requiring deep data processing. Additionally, the computer will enable work on multiple datasets for AI training, including data from banks, insurance companies, retail chains, and other sectors.
The RSC will also be used for other types of calculations. For example, it will be applied to virtual reality, evaluation of internet publications, or analysis of user behavior on social networks such as Facebook, Instagram, and Twitter.
“This opens up new possibilities for using machine learning in various sectors, from medicine to energy, to solve complex applied tasks and general-purpose challenges,” said Facebook head Mark Zuckerberg. The company expects that the solution based on RSC will accelerate the development and implementation of new technologies, such as voice translation for business conferences or in virtual reality.
How the Super-Powerful RSC Computer Was Created
The development of the computing machine began in 2020. After project preparation, engineers started assembling the first prototype. It is based on 760 NVIDIA DGX A100 systems, containing 6080 accelerator modules. Each processor node is equipped with Tensor Float tensor cores. The device is designed for 80 GB of RAM and also features PCI Express interfaces.
Even at the initial assembly stage, the supercomputer had a performance 20 times greater than that of conventional supercomputers. This allowed for accelerated work with large volumes of industrial automation data, image and video processing, and tasks related to pattern recognition and speech, as well as scientific research.
Within a year, the computer’s power will be doubled, and by the beginning of 2021, it will hold a leading position in the list of the world’s most powerful supercomputers. As of today, the computer’s performance reaches 5 exaflops. It is named the AI Research SuperCluster and was developed under the Open Compute Project (OCP). The computer is capable of performing tasks related to artificial intelligence, machine learning, big data processing, and more.
By the end of 2022, after testing is completed, Meta will present a new version of the ultra-powerful computer equipped with 16,000 NVIDIA processors. The performance of the components will allow processing data sets of up to 1 exabyte or a trillion floating-point operations per second. In comparison, the performance of Microsoft’s supercomputer based on OpenAI, which ranks fifth among the world’s most powerful computing systems, is 12 teraflops (1 exaflop = 1 million teraflops). Meta will use NVIDIA A100 technology to accelerate calculations on graphic processors.
What Is an AI Supercomputer?
The capabilities of modern computers are impressive; however, it is not entirely clear what artificial intelligence is and how it works. In theory, it is simple. It is a computing system that can learn and self-learn. It is capable not only of predicting events but also of modeling them.
For example, a supercomputer can analyze and predict the economic situation in a country and then recommend to the government what steps to take to improve it. Unlike a human, who thinks only within the framework of their profession, a supercomputer can solve problems related to other fields.
The ultra-fast computer can perform trillions of operations per second. Its computational power increases exponentially based on its size. In other words, the larger the amount of RAM, the faster it operates.
Currently, the most high-tech computer in the world is the American supercomputer Summit (OLCF-4), which has a power range of 122.3 petaflops. It operates on two IBM POWER9 processors and has a RAM volume of 250 petabytes. The computer was launched in June 2018.
What Is the Difference Between Supercomputers and AI Supercomputers?
Both types of computing technology are similar in terms of power. However, the main difference lies in the intended purpose of the machines. Multifunctional supercomputers are usually used for scientific research, such as space exploration, calculating the actions of combat installations, predicting earthquakes, etc. They are key components in most large research projects.
Machine learning does not require high precision from an AI supercomputer, while conventional supercomputers are used to solve problems with more complex algorithms. This allows AI-based supercomputers to process more operations per second. As a result, they can handle tasks that require significant computational power but do not demand high precision.