News

Tesla’s “Dojo” supercomputer, construction of which began this summer, is introduced as the world’s most powerful future supercomputer… by far. Elon Musk has announced that, with this cutting-edge equipment, he will be able to reach 100 exaflops before the end of 2024, thanks to a billion-dollar investment. An ambitious gamble: today, the most advanced computer is 100 times less powerful.

The American supercomputer Frontier passed the symbolic 1 exaflop mark for the first time in history in June 2022.

Invest twice as much to do a hundred times better? That’s Elon Musk’s challenge with “Dojo”. This supercomputer, whose construction announced by Tesla on July 19, will benefit from an investment of $1 billion over three years. The ambition: to create the world’s most powerful computer. And by far. With 100 exaflops expected for October 2024, the computer will be a 100 times more powerful than the world’s most powerful computer to date. It is destined to drive artificial intelligence models behind self-driving cars.

Surpassed for the first time in 2022 by the American computer Frontier, the exaflop barrier corresponds to the execution of 1 billion of billion floating-point operations per second. Germany and France are set to join the coveted club of exaflop supercomputers, in 2024 and 2025 respectively, with the European supercomputers Jupiter and Jules Verne. Production of the American computer began this summer, with the aim of entering the world’s top 5 at over 0.2 exaflops by January 2024. But the technical reality promises to be more complex.

A tight deadline

“Reaching 100 exaflops by the end of 2024 seems ambitious”, says Jean-Yves Berthou, Director of the Centre Inria in Saclay, France. It took Frontier almost a year to pass the exaflop mark. “Here, we would go from a situation with no machines to 100 exaflops in one year. It’s not impossible, but it’s very optimistic”, agrees François Bodin, professor at the Université de Rennes. The scientist highlights, for example, to the uncertainties surrounding the supply of GPU graphics chips, pantagrual data storage or the supply of electricity. Indeed, the average consumption of a supercomputer would be 20 megawatts, according to the CEA.

Tesla’s announcement also keeps the nature of these 100 exaflops unclear. The company has not specified whether this is a theoretical or actual capacity. The difference can be significant: corrected for factors such as memory latency or inter-node communication, Frontier achieves an actual capacity of 1.1 exaflops for a theoretical peak of 1.7 exaflops. To evaluate computer performance under real-life conditions, the supercomputers run the “Linpack” a test bench that brings together several programs and software libraries. “It’s not the same thing to 100 exaflops by running the Linpack algorithm producing 100 exaflops by running theLinpack algorithm than having a machine with a theoretical peak power producing 100 exaflops by running the Linpack algorithm than having a machine with a theoretical peak power of 100 exaflops”, sums up François Bodin.

A supercomputer like no other

Nevertheless, the major strength of Dojo’s computing power lies in its specialization. Frontier, Jupiter or Jules Verne have a wide range of applications – research in the fields of space, climate, pharmaceuticals, energy and many more. Dojo, on the other hand, has just one: training artificial intelligence models for autonomous cars. This type of calculation requires less precision than scientific calculations.

This difference enables Tesla to build a specialized processor architecture, with a lower number of bits (and therefore a lower precision of calculation). precision). “In all likelihood, they will run on 8-bit computation, rather than 64-bit as in classical scientific computing”, explains Jean-Yves Berthou. They will therefore gain a factor of 8 in power.” In other words, by agreeing to divide its calculation precision by 8, Dojo will be able to carry out 8 times as many instructions. We shouldn’t think that Europe is lagging behind, because it would invest half as much for a hundred times less performance”, argues the scientist. The exaflop metric is simply not used in the same context here.

Tesla’s gamble has already had a significant impact: the rise in its stock market price. A Morgan Stanley report estimates that Tesla’s market capitalization could rise by almost $600 billion (562 billion euros) thanks to the potential of its supercomputer, which is expected to help create robot cabs. As is often the case, Elon Musk’s announcements seduce investors. But beware : the billionaire already said he was “very confident” about the appearance of robot cabs in… 2020. His ambitions for Dojo could be just as over-optimistic.

In the era of electronic miniaturization, supercomputers are making a comeback, driven by the rise of artificial intelligence. A strategic battle is unfolding among major powers.

They were once thought to be obsolete, but supercomputers are back in action. Today, they are benefiting from a thriving global market, thanks to the exploitation of the 21st-century black gold: data for artificial intelligence (AI). They have even become a matter of sovereignty, to the extent that the United States, in its trade war with China, goes as far as denying its Asian rival microprocessors intended for these machines. In Europe, the survivors are emerging, starting with the French state, which is closely monitoring the negotiations surrounding the dismantling of the Atos Group. In its subsidiary Eviden, there is indeed a gem to be cherished, among the few that survive in Europe: supercomputers stemming from the acquisition of Bull in 2013. A new factory is even expected to be built in Angers by 2027.

These behemoths are being called upon to develop medicines and vaccines more rapidly, particularly against Covid-19, refine weather forecasts in the face of climate change, improve the aerodynamics of aircraft and other vehicles to consume less energy, combat increasingly formidable cyberattacks, or simulate a nuclear explosion in the name of deterrence. Machine learning and quantum computing (massive simultaneous calculations at the atomic scale) need them.

Long confined to academic research or industrial and nuclear simulation (defense), the lineage of supercomputers is gaining strength with artificial intelligence. “Large-scale AI models are growing very rapidly, and new buyers are starting to use high-capacity machines, with selling prices ranging from tens of millions to hundreds of millions of dollars each. Fighting against cyberattacks will also require major computing power,” predicts Earl Joseph, CEO of Hyperion Research, an American research company specializing in the global HPC (high-performance computing) market. Because the more powerful the computers, the more expensive they are.

After being acquired by the EuroHPC joint venture, this supercomputer will therefore be hosted at the end of 2025 at the CEA’s Very Large Computing Center (TGCC). It will benefit from the expertise of the latter’s High Performance Computing (HPC) division in the operation of large-scale systems such as Joliot-Curie (GENCI, for open research) and Topaze (CCRT, Center for Computing Research and Technology , for industrial research).

The main objective of the Jules Verne consortium is to deploy a world-class Exascale supercomputer, based on European hardware and software technologies. It will make it possible to respond to the major societal and scientific challenges via the convergence at the scale of digital simulations, the analysis of massive data and artificial intelligence.

Indeed, this project responds to major societal and global challenges corresponding to the national strategies of the Netherlands and France, in particular within the framework of France 2030 for the latter. The supercomputer will act as a sovereign accelerator in the finer modeling of the effects of climate change, in the development of new materials, energies and low-carbon mobility solutions, in the creation of digital twins of the human body allowing personalized medicine or still in training the next generation of generative AI or multimodal models. It will also address the challenges related to the explosion of data generated by scientific instruments (such as telescopes, satellites, sequencers, microscopes, sensor networks, etc. by IoT/Internet devices or by large simulations This avalanche of data makes the use of these supercomputers crucial for science, industry and decision-makers, in order to process this data in competitive timeframes and in the most energy-efficient way possible.

After the deployment of EuroHPC systems such as JUPITER (in Germany), the first Exascale system in Europe in 2024, Jules Verne will provide European, French and Dutch researchers with an unprecedented computing capacity of more than 1 Exaflop/s – one billion billion (“1” followed by 18 zeros) of operations per second, equivalent to over 5 million modern laptops, and over 300 PB of boot storage.

Beyond the machine itself, the Jules Verne consortium, in conjunction with other EuroHPC consortia, will provide support to European researchers for the porting and optimization of their applications on the supercomputer, as well as for training. In this perspective, the Jules Verne consortium will collaborate with all European Centers of Excellence (CoE) and end users for the implementation of the system. It has already established relationships with national R&D Exascale projects (such as the France 2030 NumPEx research program). As a reminder, the NumPEx program aims to design and develop software components that will equip future Exascale machines and prepare the major fields of scientific and industrial applications to fully exploit the capabilities of these machines. The NumPEx program has a budget of 40.8 million euros over 5 years.

The total cost of acquiring and operating the supercomputer for 5 years amounts to 542 million euros. Of this total, 271 million euros are provided by EuroHPC JU, 8 million euros by the Dutch Ministry of Culture, Education and Science and 263 million euros provided by the French Government. ONERA and IFPEN have expressed their intention to join the French part of the consortium, paving the way for other research institutes and French industrialists.

Beyond France and the Netherlands, the Jules Verne consortium is ready to welcome other countries, as partners sharing the same vision in the service of science, innovation and sovereign technologies.

EuroHPC’s approval of the Jules Verne consortium’s application is excellent news for French and European research. This is another important step in securing financing for an Exascale-class supercomputer, worth a total of €542 million.

These means of calculation will be necessary to meet the scientific and technological challenges that await us, such as climate change, energy transition or health. The supercomputer will therefore play a key role in guaranteeing our technological sovereignty and our industrial competitiveness, and I hope that new public and private partners will join the consortium in the coming weeks.

Sylvie Retailleau, French Minister for Higher Education and Research

It is excellent news that the European scientific community, with France and the Netherlands in the lead, is joining forces to produce the supercomputer proposed by the Jules Verne consortium. Europe is thus reaffirming its position on the global research scene. Author Jules Verne has piqued our curiosity with stories about a technological future where one can travel to the Moon or the deep sea. Thanks to this supercomputer, we are doing it again. With this immense computing power, scientists have a glimpse of the future, allowing them to help solve fundamental societal problems in areas such as health or the fight against climate change.

Robbert Dijkgraaf, Dutch Minister of Education, Culture and Science

A billion billion operations per second to accelerate the advent of the future. GENCI is delighted with the announcement by EuroHPC of the selection of the Franco-Dutch consortium Jules Verne to host and operate an Exascale class supercomputer. It is an international recognition of French scientific and technical expertise in combining the applications of digital simulation, massive data analysis, artificial intelligence and soon hybrid quantum computing and by implementing hardware and European software.

Above all, these are the first steps in the era of the Exascale which will allow our national research communities to realize the dream of simulating complex phenomena to solve historical scientific puzzles as well as the possibility of being able to be creative in devices to meet the industrial and societal challenges of energy, innovative materials and health, such as the treatment of neurodegenerative diseases.

Philippe Lavocat, CEO of GENCI

This supercomputer will be an exceptional instrument for European research at the service of European society and sovereignty. It will enable major advances in many fields that are at the heart of CEA’s research activities, such as high-resolution climate modelling, fusion for energy, innovative materials, human digital twins and personalized medicine. It will provide our researchers and industrialists with world-class computing resources to exploit the deluges of data linked to the deployment of new digital systems, autonomously, and thus remain in the global race. The CEA has a long experience of designing and implementing pre-exascale supercomputers in state-of-the-art computing centers.

We will put all our expertise in the design and operation of computing centers at the service of this project, with the objective of performance and control of energy consumption.

François Jacq, General Administrator of the CEA

We are proud to work together in the Jules Verne Consortium to significantly advance research on societal challenges. This supercomputer will help Dutch researchers perform complex simulations in areas ranging from climate science to medicine and astronomy. We are very proud that our experts in large-scale computing, SURF, can contribute to this and thus help researchers in their work.

Jet de Ranitz, CEO of SURF

The Digital Exploratory Exascale Priority Research Program and Equipment (PEPR) – piloted by the CNRS, the CEA and Inria – aims to design and develop software bricks that will equip future exascale machines. It thus contributes to preparing future users, scientists and industrialists, to exploit the capabilities of these machines. The program has a budget of 40.8 million euros over 8 years. Explanations with Michel Daydé, co-director of the program for the CNRS.

Notes

1. A group of programs that work together to produce a result or achieve a common goal.