Atos Tees Up New BullSequana Supercomputer for European Exascale

By Tiffany Trader

February 16, 2022

At a three-hour livestreamed launch event held today (Feb. 16) in Paris, French technology firm Atos unveiled its new supercomputer, the BullSequana XH3000, intended to provide from one petaflops to exaflops of traditional double-precision digital simulation and up to 10 exaflops of mixed-precision AI performance. Availability is planned for the fourth quarter of this year.

Few technical details were disclosed, but Atos said the XH3000 will support the latest processing components from AMD, Intel and Nvidia, including the forthcoming Grace Arm CPU. The architecture will also support the future Rhea processor that SiPearl is designing on behalf of the European Processor Initiative. As detailed in a presentation (PDF) from November, the XH3000 has standardized slots for compute and/or switch blade modules and supports up to 38 blades per rack.

Concept QPU chip. Source: Atos.

As part of today’s reveal, Atos teased a forthcoming QPU component, which will provide quantum acceleration and is an extension of Atos’ Quantum Learning Machine (Atos QLM) project. “The QPU that we’ll be able to integrate into our machine, that would be probably one of the first on the planet to do that,” said Agnès Boudot, head of HPC and quantum at Atos.

Further, Boudot noted, an effort is underway to bridge traditional on-prem HPC with cloud environments leveraging the JARVICE XE HPC cloud technology, recently acquired with the purchase of Nimbix.

Atos estimates the XH3000 will draw on the order of 20MW at full scale and says the system is twice as energy-efficient as its previous generation machine, the XH2000. Atos further touts a six-fold increase in power density, which it claims constitutes “the best computing power per m² on the market.” The increased density will enable the XH3000 to support chips that consume up to (or even above) 1000 watts, versus ~350 watts for the current-generation XH2000 system.

The XH3000 features Atos’ fourth-generation direct liquid cooling (DLC) technology that uses 40°C inlet water. The patented cooling system provides over 50 percent more cooling power than previous generations, according to Atos, which cited an eco-compliant product lifecycle that spans “raw material sourcing and manufacturing, to testing and transportation, and disposal/recycling.”

The new system will also support a larger number of interconnect options, including Atos’ BXI technology, high-speed Ethernet, and HDR and NDR InfiniBand from Nvidia. The interconnect midplane of the XH2000-generation BullSequana is not capable of NDR support.

Agnes Boudot shows off the new Atos BullSequana system

Boudot led the unveiling ceremony for roughly 80 in-person attendees and a global audience of virtual viewers. The XH3000 cabinet is recognizably Atos in design, embracing the same modern, chrome aesthetic as previous generations while introducing a unique Turing pattern symbolizing the living elements of the planet. “It’s close to nature,” said Boudot, “and it is also an idea for us to show the path to decarbonization and the importance of [sustainability],” touching on one of the major themes for the launch event, along with hybrid technologies and European sovereignty.

Opening the outer doors to display the inner components, Boudot said, “It looks very simple, but it’s a lot of IP. It’s a lot of innovation … the result of 15 years of experience from Bull and Atos.”

On the naming convention for the XH3000, Boudot explained that the X is for exascale and the H is for hybrid, although we note the same letters are in use for the current-generation XH2000.

The XH3000 is the follow-on to the successful XH2000 design, which itself is the foundation for a number of EuroHPC systems: Leonardo at Cineca in Italy; Discoverer at the SOFIA Tech Park in Bulgaria; Vega at IZUM in Slovenia; MeluXina in Luxembourg; and Deucalion in Portugal (via a collaboration with Fujitsu). Atos also supplied Europe’s fastest supercomputer to Forschungszentrum Jülich in Germany with the delivery of the JUWELS Booster in 2020.

Atos has aggregated 40 systems across the Top500 list. “The supercomputers we deployed will contribute to European scientific research capabilities,” said Atos CEO Rodolphe Belmer at the launch event. “They will be used for the development of scientific, public and industrial applications, including bioinformatics, pharmaceuticals, molecular dynamics and mechanics, meteorology and the fight against climate change.” He emphasized that all 40 machines also appear on the Green500 list, ranked for energy-efficiency. MeluXina, one of eight new EuroHPC systems, is the ninth-greenest system, delivering 26.96 gigaflops-per-watt.

“[The XH3000] is Atos’ most efficient and powerful supercomputer and an important element in securing today’s digital and economic sovereignty,” said Belmer.

Nearly everyone who spoke at the launch event expressed support for strengthening Europe’s digital sovereignty and viewed the Atos BullSequana XH3000 as a critical part of the path to achieving European supercomputing independence, or at least partial independence. It was acknowledged by Jean-Philippe Nominé, a fellow and HPC strategic collaborations manager with the simulation and information sciences division at CEA, that the EPI/SiPearl processor would most likely be manufactured in Asia, at least initially. There is now the possibility to one day have the foundry in Europe, said Nominé, citing the possibilities opened up by the new European Chips Act.

Nominé went on to emphasize that sovereignty of applications absolutely matters. “It’s very important to have the control of what we do in the applications, the software as well. And to keep control of both of these for our purposes, and also to have the knowledge and skills which are necessary to use the application in the proper way,” said Nominé.

During a roundtable panel focused on the role of extreme computing for industrial innovation and research projects, moderator Addison Snell, CEO of HPC analyst firm Intersect360 Research, remarked that Atos is “essentially the vendor face of European supercomputing.”

With the launch of BullSequana XH3000, Atos says it is on track to deliver exascale systems within the next two years. The EuroHPC consortium plans to stand up two exascale systems by the end of 2024, with France and Germany each intending to host one of the systems. No formal system selection has been made yet, but Atos is certainly a major contender for the contracts.

A draft timeline of EuroHPC’s activities over the next six years (click to expand).

Après le exascale

After exascale comes zettascale, said Arnaud Bertrand, senior vice president and head of big data and security strategy, innovation and R&D at Atos, providing the closing remarks for the launch event.

“The zettascale era won’t be like exascale,” said Bertrand. “We will see many changes in the paradigm, and probably even in the way we have been designing computers for the last 70 years. My bet is that zettascale will open the era of composable architecture. This will require a change compared to today’s paradigm, address[ing] challenges in four areas.” The four elements coming together in Bertrand’s vision of zettascale are: aggregation with multiple quantum accelerators, the marriage of AI and simulation, the advance of the smart network that will compute on the data while transporting it, and finally, 100 percent carbon neutrality.

“Of course, we won’t create an HPC [system] which is not consuming electricity and is just powered by the wind,” continued Bertrand. “That’s not possible today. Not yet. It will be a huge challenge [to achieve] 100 percent carbon neutrality, [but] we are working on it.”

Update:

Initial technical specs provided by Atos for the BullSequana XH3000 architecture stated that it scales to one exaflops. Reached for clarification, Eric Eppe (head of portfolio and solutions for HPC and quantum at Atos) said, “[The architecture] has been engineered to achieve, at least, one exaflops. The XH3000 is built to scale to hundreds of racks, the sizing being determined by the technology (CPU/GPU/IPU) that powers it. The more powerful or efficient the technology is, the smaller the system will need to be. The accepted range for an exaflopic system in 2023/2025 is around 10,000 nodes and 25,000 endpoints. The BullSequana XH3000 can be below or above this number depending on the technology that is put in it.”

To learn more about the XH3000 launch event, read the additional feature coverage from HPCwire here.

This story has been updated to provide details about the scalability of the system.

Subscribe to HPCwire's Weekly Update!

Be the most informed person in the room! Stay ahead of the tech trends with industry updates delivered to you every week!

Quantum Companies D-Wave and Rigetti Again Face Stock Delisting

October 4, 2024

Both D-Wave (NYSE: QBTS) and Rigetti (Nasdaq: RGTI) are again facing stock delisting. This is a third time for D-Wave, which issued a press release today following notification by the SEC. Rigetti was notified of delisti Read more…

Alps Scientific Symposium Highlights AI’s Role in Tackling Science’s Biggest Challenges

October 4, 2024

ETH Zürich recently celebrated the launch of the AI-optimized “Alps” supercomputer with a scientific symposium focused on the future possibilities of scientific AI thanks to increased compute power and a flexible ar Read more…

The New MLPerf Storage Benchmark Runs Without ML Accelerators

October 3, 2024

MLCommons is known for its independent Machine Learning (ML) benchmarks. These benchmarks have focused on mathematical ML operations and accelerators (e.g., Nvidia GPUs). Recently, MLCommons introduced the results of its Read more…

DataPelago Unveils Universal Engine to Unite Big Data, Advanced Analytics, HPC, and AI Workloads

October 3, 2024

DataPelago today emerged from stealth with a new virtualization layer that it says will allow users to move AI, data analytics, and ETL workloads to whatever physical processor they want, without making code changes, the Read more…

IBM Quantum Summit Evolves into Developer Conference

October 2, 2024

Instead of its usual quantum summit this year, IBM will hold its first IBM Quantum Developer Conference which the company is calling, “an exclusive, first-of-its-kind.” It’s planned as an in-person conference at th Read more…

Stayin’ Alive: Intel’s Falcon Shores GPU Will Survive Restructuring

October 2, 2024

Intel's upcoming Falcon Shores GPU will survive the brutal cost-cutting measures as part of its "next phase of transformation." An Intel spokeswoman confirmed that the company will release Falcon Shores as a GPU. The com Read more…

The New MLPerf Storage Benchmark Runs Without ML Accelerators

October 3, 2024

MLCommons is known for its independent Machine Learning (ML) benchmarks. These benchmarks have focused on mathematical ML operations and accelerators (e.g., Nvi Read more…

DataPelago Unveils Universal Engine to Unite Big Data, Advanced Analytics, HPC, and AI Workloads

October 3, 2024

DataPelago today emerged from stealth with a new virtualization layer that it says will allow users to move AI, data analytics, and ETL workloads to whatever ph Read more…

Stayin’ Alive: Intel’s Falcon Shores GPU Will Survive Restructuring

October 2, 2024

Intel's upcoming Falcon Shores GPU will survive the brutal cost-cutting measures as part of its "next phase of transformation." An Intel spokeswoman confirmed t Read more…

How GenAI Will Impact Jobs In the Real World

September 30, 2024

There’s been a lot of fear, uncertainty, and doubt (FUD) about the potential for generative AI to take people’s jobs. The capability of large language model Read more…

IBM and NASA Launch Open-Source AI Model for Advanced Climate and Weather Research

September 25, 2024

IBM and NASA have developed a new AI foundation model for a wide range of climate and weather applications, with contributions from the Department of Energy’s Read more…

Intel Customizing Granite Rapids Server Chips for Nvidia GPUs

September 25, 2024

Intel is now customizing its latest Xeon 6 server chips for use with Nvidia's GPUs that dominate the AI landscape. The chipmaker's new Xeon 6 chips, also called Read more…

Building the Quantum Economy — Chicago Style

September 24, 2024

Will there be regional winner in the global quantum economy sweepstakes? With visions of Silicon Valley’s iconic success in electronics and Boston/Cambridge� Read more…

How GPUs Are Embedded in the HPC Landscape

September 23, 2024

Grasping the basics of Graphics Processing Unit (GPU) architecture is crucial for understanding how these powerful processors function, particularly in high-per Read more…

Shutterstock_2176157037

Intel’s Falcon Shores Future Looks Bleak as It Concedes AI Training to GPU Rivals

September 17, 2024

Intel's Falcon Shores future looks bleak as it concedes AI training to GPU rivals On Monday, Intel sent a letter to employees detailing its comeback plan after Read more…

Nvidia Shipped 3.76 Million Data-center GPUs in 2023, According to Study

June 10, 2024

Nvidia had an explosive 2023 in data-center GPU shipments, which totaled roughly 3.76 million units, according to a study conducted by semiconductor analyst fir Read more…

Granite Rapids HPC Benchmarks: I’m Thinking Intel Is Back (Updated)

September 25, 2024

Waiting is the hardest part. In the fall of 2023, HPCwire wrote about the new diverging Xeon processor strategy from Intel. Instead of a on-size-fits all approa Read more…

AMD Clears Up Messy GPU Roadmap, Upgrades Chips Annually

June 3, 2024

In the world of AI, there's a desperate search for an alternative to Nvidia's GPUs, and AMD is stepping up to the plate. AMD detailed its updated GPU roadmap, w Read more…

Ansys Fluent® Adds AMD Instinct™ MI200 and MI300 Acceleration to Power CFD Simulations

September 23, 2024

Ansys Fluent® is well-known in the commercial computational fluid dynamics (CFD) space and is praised for its versatility as a general-purpose solver. Its impr Read more…

Shutterstock_1687123447

Nvidia Economics: Make $5-$7 for Every $1 Spent on GPUs

June 30, 2024

Nvidia is saying that companies could make $5 to $7 for every $1 invested in GPUs over a four-year period. Customers are investing billions in new Nvidia hardwa Read more…

Shutterstock 1024337068

Researchers Benchmark Nvidia’s GH200 Supercomputing Chips

September 4, 2024

Nvidia is putting its GH200 chips in European supercomputers, and researchers are getting their hands on those systems and releasing research papers with perfor Read more…

Comparing NVIDIA A100 and NVIDIA L40S: Which GPU is Ideal for AI and Graphics-Intensive Workloads?

October 30, 2023

With long lead times for the NVIDIA H100 and A100 GPUs, many organizations are looking at the new NVIDIA L40S GPU, which it’s a new GPU optimized for AI and g Read more…

Leading Solution Providers

Contributors

Everyone Except Nvidia Forms Ultra Accelerator Link (UALink) Consortium

May 30, 2024

Consider the GPU. An island of SIMD greatness that makes light work of matrix math. Originally designed to rapidly paint dots on a computer monitor, it was then Read more…

IBM Develops New Quantum Benchmarking Tool — Benchpress

September 26, 2024

Benchmarking is an important topic in quantum computing. There’s consensus it’s needed but opinions vary widely on how to go about it. Last week, IBM introd Read more…

Quantum and AI: Navigating the Resource Challenge

September 18, 2024

Rapid advancements in quantum computing are bringing a new era of technological possibilities. However, as quantum technology progresses, there are growing conc Read more…

Intel Customizing Granite Rapids Server Chips for Nvidia GPUs

September 25, 2024

Intel is now customizing its latest Xeon 6 server chips for use with Nvidia's GPUs that dominate the AI landscape. The chipmaker's new Xeon 6 chips, also called Read more…

Google’s DataGemma Tackles AI Hallucination

September 18, 2024

The rapid evolution of large language models (LLMs) has fueled significant advancement in AI, enabling these systems to analyze text, generate summaries, sugges Read more…

Microsoft, Quantinuum Use Hybrid Workflow to Simulate Catalyst

September 13, 2024

Microsoft and Quantinuum reported the ability to create 12 logical qubits on Quantinuum's H2 trapped ion system this week and also reported using two logical qu Read more…

IonQ Plots Path to Commercial (Quantum) Advantage

July 2, 2024

IonQ, the trapped ion quantum computing specialist, delivered a progress report last week firming up 2024/25 product goals and reviewing its technology roadmap. Read more…

US Implements Controls on Quantum Computing and other Technologies

September 27, 2024

Yesterday the Commerce Department announced export controls on quantum computing technologies as well as new controls for advanced semiconductors and additive Read more…

  • arrow
  • Click Here for More Headlines
  • arrow
HPCwire