As LUMI Launches, a Look at the State of EuroHPC

By Oliver Peckham

June 13, 2022

Today, the LUMI pre-exascale supercomputer was inaugurated in Kajaani, Finland. LUMI—which currently weighs in around 152 Linpack petaflops, but is expected to soon exceed 375—represents the largest success thus far of the EuroHPC Joint Undertaking, Europe’s concerted supercomputing play. LUMI marks the beginning of the end for the first phase of EuroHPC, but the second phase (keywords: “exascale” and “quantum”) is already underway. Here, we cover the state of EuroHPC systems and initiatives both past and future, informed by both an HPCwire interview with Anders Jensen, executive director of EuroHPC, and his session at ISC 2022.

Already delivered

In June 2019, EuroHPC selected eight supercomputer centers across the European Union to host its first eight systems—five decidedly petascale systems, and three “pre-exascale” systems designed to rate in the hundreds of petaflops. Five of those octuplets have already been delivered.

Vega, the first EuroHPC supercomputer delivered, is a 1,020-node Atos-built system with AMD Epyc “Rome” CPUs and Nvidia A100 GPUs that launched in Slovenia in April 2021. The system has two modules listed on the spring Top500 list: its 960-node CPU partition (3.82 Linpack petaflops, #131) and its 60-node GPU partition (3.10 Linpack petaflops, #172).

MeluXina is an 813-node Atos-built system with AMD Epyc “Rome” CPUs and Nvidia A100 GPUs that launched in Luxembourg in June 2021. MeluXina also has two modules listed on the spring Top500 list: its 573-node cluster module (2.29 Linpack petaflops, #306) and its 200-node accelerator module (10.52 Linpack petaflops, #48).

Karolina is an 831-node HPE-built system with AMD Epyc “Rome” CPUs and Nvidia A100 GPUs (along with a handful of Intel Xeon CPUs) that launched in Czechia in the summer of 2021. Karolina, too, has two modules listed on the spring Top500 list: its CPU partition (2.84 Linpack petaflops, #202) and its GPU partition (6.75 Linpack petaflops, #79).

Discoverer is a 1,128-node Atos-built system with AMD Epyc “Rome” CPUs that launched in Bulgaria in October 2021. At 4.52 Linpack petaflops, the system ranked 113th on the spring Top500 list.

LUMI, the first of EuroHPC’s pre-exascale systems to be delivered, is a 4,112-node HPE-built system with AMD Epyc “Trento” CPUs and AMD MI250X GPUs (the same architecture as the new Frontier exascale supercomputer) that launched in Finland today. Currently, LUMI’s main 2,560-node GPU-accelerated partition (LUMI-G) rates at 151.90 Linpack petaflops, earning third place on the spring Top500 list, while its 1,536-node CPU partition (LUMI-C) rates at 6.30 Linpack petaflops and placed 84th. LUMI’s power is expected to dramatically increase in the coming weeks, eventually exceeding 375 Linpack petaflops.

The LUMI supercomputer. Image courtesy of CSC.

Later this year

Deucalion will be the last of EuroHPC’s first five petascale systems when it is delivered in Portugal later this year. Deucalion, interestingly, will consist of three partitions: a 1,632-node Arm partition based on Fujitsu’s A64FX processor (expected to deliver 3.8 Linpack petaflops), a 500-node x86 partition powered by AMD Epyc “Rome” CPUs (1.62 Linpack petaflops) and a 33-node accelerated partition with both AMD CPUs and Nvidia A100 GPUs. “[For] Deucalion, we’re … waiting for the datacenter to be finalized, everything else is ready,” Jensen said.

Leonardo will be the second of EuroHPC’s pre-exascale systems when it arrives later this year in Italy. The Atos-built supercomputer will include a 1,536-node data-centric module with Intel Xeon “Sapphire Rapids” CPUs and a 3,456-node booster module with Intel Xeon “Ice Lake” CPUs and Nvidia A100 GPUs (240.5 Linpack petaflops). “Leonardo … is being constructed as we sit here and I hope and expect that we will see an announcement on the November Top500,” Jensen said.

The problem child

MareNostrum 5, the long-awaited successor to the MareNostrum 4 system at the Barcelona Supercomputing Center (BSC), is EuroHPC’s problem child. In October 2020, Jensen had said to expect an announcement on the procurement “in the coming weeks.” That did not come to pass; the tender was canceled in May 2021 amid much purported drama before being reissued in late December. That second tender closed at the end of January 2022.

“[For] MareNostrum 5, we went through the process of the procurement, we’ve closed that, and with a little luck, within the next week or so you’ll be able to see an announcement on that moving forward and with a timeline,” Jensen said. “We’re in the final rounds of contract negotiations.” (It has now been two weeks since that interview, and we eagerly await further news on MareNostrum 5.)

MareNostrum 5 is aimed at delivering at least 205 Linpack petaflops.

An outline of all eight systems in EuroHPC’s first slate, courtesy of the Joint Undertaking.

The future

All eight of those systems are covered by EuroHPC’s first regulation from the European Commission. That first regulation covered the slate through 2021; now, the Joint Undertaking has a new regulation—and a new budget—for the coming six years. That budget (around €7 billion) consists of around €3.5 billion from the Digital Europe Program, the Horizon Europe Program and the Connecting Europe Facility, all of which is expected to be matched by national contributions from EuroHPC member states that host forthcoming systems and facilities.

Europe’s first exascale systems (and more!)

Speaking of which: included in that budget are at least two exascale systems, with the site selection for one coming very, very soon. “Shortly before Christmas, we went out with that call—which countries would like to host essentially the first European exascale system?” Jensen said. “And we’ve had the responses back, we’ve evaluated them; now, I can’t disclose any more than that other than that it’s on the agenda of our governing board meeting, which takes place in Kajaani—we’ve colocated it with the inauguration of LUMI. So, on the 14th of June, the governing board is to decide who will host Europe’s first exascale machine along with also a number of new mid-ranges.”

“It’s clear we’re going to continue to invest in machines, but it’s also for geographical diversity—we want to get more users on it, we want to increase the knowledge within HPC,” he added. “So I think you’re going to continue to see new countries joining the HPC scene and wishing to have machines of their own so they can grow the ecosystem within their country.”

As far as when we can expect to learn the outcomes of that meeting: “I think you can expect some announcements shortly following that [meeting],” Jensen said, “and then starts the hard work of actually delivering.”

Targeted aggregate European computing power through 2024, including two exascale systems and new petascale systems. Image courtesy of Anders Jensen.

Quantum

Quantum computing is also part of EuroHPC’s second phase—with an asterisk. “With the new regulation, quantum was added to the EuroHPC agenda, and it was added heavily,” Jensen said. “So we’ve gone out again with a call where we’re asking who wishes to host the first EuroHPC quantum computers. That call has not yet closed.” Jensen did, during his session on EuroHPC at ISC 2022, clarify that much of the research and innovation on quantum technologies still rests under the purview of the European Commission; as it stands, EuroHPC has been tasked with delivering quantum infrastructure for scientists and coupling quantum technologies to HPC.

“The aim is to provide access to quantum computers to the European community just like we’re providing access to HPC systems,” he said. “We are focusing on diversity in technology, so what we would like to have is as many quantum technologies available so that scientists and any potential users of them can get to know what using quantum is all about and we can enrich our knowledge on that.”

The Joint Undertaking, he said, was also looking to federate and interconnect its existing machines. “Right now, we’re buying a number of systems and we’re making them available, but we can’t help but think that putting them together, some of them, into bigger systems, could have potential,” Jensen said. And vis-a-vis federation: as the number of EuroHPC systems grows, Jensen said, “we can’t keep on managing them as individual machines.”

The far future

“The regulation also dictates post-exascale, but we have yet to define what post-exascale means,” Jensen said.

Homegrown talent and tech

Jensen also stressed the importance of building HPC talent and technology in Europe. “I think the whole Covid situation and … different geopolitical situations have led Europe to wake up to the fact—and it’s not just in HPC, it’s just in general—that we’ve become very reliant on other parts of the world supplying what we may not realize [are] very critical components,” Jensen said. “And then all of a sudden you need it, and that’s when you realize it’s critical.”

“We have a very ambitious agenda of wanting a significant European technology footprint within the second exascale system that we’re going to fund,” Jensen continued; in his talk the following day, he specified a processor built by the European Processor Initiative (EPI). Like EuroHPC, EPI entered its second phase this year; that second phase aims to finalize EPI’s first-gen low-power processors and develop second-generation processors. “In parallel to that, you have the whole [European Chips Act] that’s come to life via the Commission, and all the interesting things that that’s going to bring,” Jensen continued, “and on the EuroHPC side we have already a placeholder for research and innovation calls in and around RISC-V and other processor technologies.”

“Europe needs to regain some self-sufficiency—some ability to do things ourselves,” Jensen said. “That doesn’t mean we don’t want to play with others, it just means we also want to be able to do things. And on the technology side, something like a European processor is an extremely important component in order to be able to rest assured … [that] we would be more self-sustaining. But also, it’s not just about that—it’s also about regaining that knowledge of how to do it so that we have the expertise and we can build an ecosystem around it.”

To that end, Jensen said that EuroHPC will also be issuing “a number of calls to further increase the knowledge of HPC and the skills within HPC.”

Alongside that emphasis on homegrown European technology, non-European companies like HPE and Intel have also been investing heavily in establishing fabs, factories and innovation centers in Europe in recent months.

“What we’ve seen from a number of players is a willingness to play with Europe, to be part of this,” Jensen said when asked about these developments. “What we’re saying as Europe is: we want to be part of it! We don’t want just to buy something off the shelf that we ship in—no, we want to be part of the process. And I think what a number of manufacturers and non-European companies have seen is: there is some expertise within Europe that they can also tap into.”

“We’ve clearly seen a change in a number of non-European players, which I really welcome,” he added, “and I’d like to think that the engagement of EuroHPC had at least a little to do with it.”

Header image: LUMI’s new datacenter. Image courtesy of CSC.

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!

MLPerf Inference 4.0 Results Showcase GenAI; Nvidia Still Dominates

March 28, 2024

There were no startling surprises in the latest MLPerf Inference benchmark (4.0) results released yesterday. Two new workloads — Llama 2 and Stable Diffusion XL — were added to the benchmark suite as MLPerf continues Read more…

Q&A with Nvidia’s Chief of DGX Systems on the DGX-GB200 Rack-scale System

March 27, 2024

Pictures of Nvidia's new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on social media for the sheer amount of computing power it brings to artificial intelligence.  Nvidia's DGX Read more…

Call for Participation in Workshop on Potential NSF CISE Quantum Initiative

March 26, 2024

Editor’s Note: Next month there will be a workshop to discuss what a quantum initiative led by NSF’s Computer, Information Science and Engineering (CISE) directorate could entail. The details are posted below in a Ca Read more…

Waseda U. Researchers Reports New Quantum Algorithm for Speeding Optimization

March 25, 2024

Optimization problems cover a wide range of applications and are often cited as good candidates for quantum computing. However, the execution time for constrained combinatorial optimization applications on quantum device Read more…

NVLink: Faster Interconnects and Switches to Help Relieve Data Bottlenecks

March 25, 2024

Nvidia’s new Blackwell architecture may have stolen the show this week at the GPU Technology Conference in San Jose, California. But an emerging bottleneck at the network layer threatens to make bigger and brawnier pro Read more…

Who is David Blackwell?

March 22, 2024

During GTC24, co-founder and president of NVIDIA Jensen Huang unveiled the Blackwell GPU. This GPU itself is heavily optimized for AI work, boasting 192GB of HBM3E memory as well as the the ability to train 1 trillion pa Read more…

MLPerf Inference 4.0 Results Showcase GenAI; Nvidia Still Dominates

March 28, 2024

There were no startling surprises in the latest MLPerf Inference benchmark (4.0) results released yesterday. Two new workloads — Llama 2 and Stable Diffusion Read more…

Q&A with Nvidia’s Chief of DGX Systems on the DGX-GB200 Rack-scale System

March 27, 2024

Pictures of Nvidia's new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on social media for the sheer amount of computing po Read more…

NVLink: Faster Interconnects and Switches to Help Relieve Data Bottlenecks

March 25, 2024

Nvidia’s new Blackwell architecture may have stolen the show this week at the GPU Technology Conference in San Jose, California. But an emerging bottleneck at Read more…

Who is David Blackwell?

March 22, 2024

During GTC24, co-founder and president of NVIDIA Jensen Huang unveiled the Blackwell GPU. This GPU itself is heavily optimized for AI work, boasting 192GB of HB Read more…

Nvidia Looks to Accelerate GenAI Adoption with NIM

March 19, 2024

Today at the GPU Technology Conference, Nvidia launched a new offering aimed at helping customers quickly deploy their generative AI applications in a secure, s Read more…

The Generative AI Future Is Now, Nvidia’s Huang Says

March 19, 2024

We are in the early days of a transformative shift in how business gets done thanks to the advent of generative AI, according to Nvidia CEO and cofounder Jensen Read more…

Nvidia’s New Blackwell GPU Can Train AI Models with Trillions of Parameters

March 18, 2024

Nvidia's latest and fastest GPU, codenamed Blackwell, is here and will underpin the company's AI plans this year. The chip offers performance improvements from Read more…

Nvidia Showcases Quantum Cloud, Expanding Quantum Portfolio at GTC24

March 18, 2024

Nvidia’s barrage of quantum news at GTC24 this week includes new products, signature collaborations, and a new Nvidia Quantum Cloud for quantum developers. Wh Read more…

Alibaba Shuts Down its Quantum Computing Effort

November 30, 2023

In case you missed it, China’s e-commerce giant Alibaba has shut down its quantum computing research effort. It’s not entirely clear what drove the change. Read more…

Nvidia H100: Are 550,000 GPUs Enough for This Year?

August 17, 2023

The GPU Squeeze continues to place a premium on Nvidia H100 GPUs. In a recent Financial Times article, Nvidia reports that it expects to ship 550,000 of its lat Read more…

Shutterstock 1285747942

AMD’s Horsepower-packed MI300X GPU Beats Nvidia’s Upcoming H200

December 7, 2023

AMD and Nvidia are locked in an AI performance battle – much like the gaming GPU performance clash the companies have waged for decades. AMD has claimed it Read more…

DoD Takes a Long View of Quantum Computing

December 19, 2023

Given the large sums tied to expensive weapon systems – think $100-million-plus per F-35 fighter – it’s easy to forget the U.S. Department of Defense is a Read more…

Synopsys Eats Ansys: Does HPC Get Indigestion?

February 8, 2024

Recently, it was announced that Synopsys is buying HPC tool developer Ansys. Started in Pittsburgh, Pa., in 1970 as Swanson Analysis Systems, Inc. (SASI) by John Swanson (and eventually renamed), Ansys serves the CAE (Computer Aided Engineering)/multiphysics engineering simulation market. Read more…

Choosing the Right GPU for LLM Inference and Training

December 11, 2023

Accelerating the training and inference processes of deep learning models is crucial for unleashing their true potential and NVIDIA GPUs have emerged as a game- Read more…

Intel’s Server and PC Chip Development Will Blur After 2025

January 15, 2024

Intel's dealing with much more than chip rivals breathing down its neck; it is simultaneously integrating a bevy of new technologies such as chiplets, artificia Read more…

Baidu Exits Quantum, Closely Following Alibaba’s Earlier Move

January 5, 2024

Reuters reported this week that Baidu, China’s giant e-commerce and services provider, is exiting the quantum computing development arena. Reuters reported � Read more…

Leading Solution Providers

Contributors

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…

Shutterstock 1179408610

Google Addresses the Mysteries of Its Hypercomputer 

December 28, 2023

When Google launched its Hypercomputer earlier this month (December 2023), the first reaction was, "Say what?" It turns out that the Hypercomputer is Google's t Read more…

AMD MI3000A

How AMD May Get Across the CUDA Moat

October 5, 2023

When discussing GenAI, the term "GPU" almost always enters the conversation and the topic often moves toward performance and access. Interestingly, the word "GPU" is assumed to mean "Nvidia" products. (As an aside, the popular Nvidia hardware used in GenAI are not technically... Read more…

Shutterstock 1606064203

Meta’s Zuckerberg Puts Its AI Future in the Hands of 600,000 GPUs

January 25, 2024

In under two minutes, Meta's CEO, Mark Zuckerberg, laid out the company's AI plans, which included a plan to build an artificial intelligence system with the eq Read more…

Google Introduces ‘Hypercomputer’ to Its AI Infrastructure

December 11, 2023

Google ran out of monikers to describe its new AI system released on December 7. Supercomputer perhaps wasn't an apt description, so it settled on Hypercomputer Read more…

China Is All In on a RISC-V Future

January 8, 2024

The state of RISC-V in China was discussed in a recent report released by the Jamestown Foundation, a Washington, D.C.-based think tank. The report, entitled "E Read more…

Intel Won’t Have a Xeon Max Chip with New Emerald Rapids CPU

December 14, 2023

As expected, Intel officially announced its 5th generation Xeon server chips codenamed Emerald Rapids at an event in New York City, where the focus was really o Read more…

IBM Quantum Summit: Two New QPUs, Upgraded Qiskit, 10-year Roadmap and More

December 4, 2023

IBM kicks off its annual Quantum Summit today and will announce a broad range of advances including its much-anticipated 1121-qubit Condor QPU, a smaller 133-qu Read more…

  • arrow
  • Click Here for More Headlines
  • arrow
HPCwire