{"id":84628,"date":"2023-06-26T11:15:00","date_gmt":"2023-06-26T09:15:00","guid":{"rendered":"https:\/\/aktuelles.uni-frankfurt.de\/?p=84628"},"modified":"2025-07-03T12:15:39","modified_gmt":"2025-07-03T10:15:39","slug":"problem-solver-with-quantum-booster","status":"publish","type":"post","link":"https:\/\/aktuelles.uni-frankfurt.de\/en\/english\/problem-solver-with-quantum-booster\/","title":{"rendered":"Problem solver with quantum booster"},"content":{"rendered":"

How supercomputers of the future will look<\/h4>\n\n\n\n
<\/div>\n\n\n\n
\n
\"\"<\/span>
\n

<\/p>\n<\/div><\/div>\n<\/div>\n\n\n\n

<\/div>\n\n\n\n
\n
\n

The Center for Scientific Computing at Goethe University Frankfurt is currently operating a cluster with 25,000 CPU cores (processors) and 880 AMD MI50 GPUs (graphics processing units). The CPU performance is two petaflops per second, the GPU performance is six petaflops per second. A petaflop is equivalent to one quadrillion operations per second. An upgrade is planned for 2023, the procurement process is underway. Lippert: \u201cOur expansion philosophy is clear: we will continuously renew the high-performance system in line with the needs of research and maximum energy efficiency, instead of completely replacing the system in cycles of four to five years, as is otherwise usual with supercomputers.\u201d Photo: J\u00fcrgen Lecher<\/p>\n<\/div>\n<\/div>\n\n\n\n

<\/div>\n\n\n\n

When many processors are grouped together to form a supercomputer, we call this a high-performance computer. Goethe University Frankfurt has such a computer cluster and will soon be adding a quantum computer, with the perspective of tackling computing tasks which are currently unsolvable.<\/strong><\/p>\n\n\n\n

<\/div>\n\n\n\n

Leonardo, Fugaku and JUPITER are their names, and they have enormous computing power. They are all supercomputers, rack upon rack of IT cabinets housed in bare rooms in university buildings, flashing mysteriously while doing their gigantic computing jobs. Together, they form a high-performance computing cluster. In the world of high-performance computing, new records are set on a regular basis. The current No. 1 on the Top500 list, the Frontier supercomputer at the Oak Ridge National Laboratory in the US, was the first to officially break the legendary 1 exaflop speed barrier. An exaflop is one quintillion \u2013 1\u2009000 000\u2009000\u2009000\u2009000\u2009000 \u2013 computing operations per second (or one trillion in the European numbering system). Some of the most complex computations can be performed with this immense processing power: climate models that reach well into the 21st century, or simulations of the spatial structures that the coronavirus spike protein is able to assume. Supercomputers are often set to work on the big issues. How can we save the climate? How can we fight pandemics? However, the complexity of the tasks continues to grow, which is why it takes more than just ever more operations per second. High-performance computing needs new ideas for using the combined processing power of several supercomputers in a smart way. <\/p>\n\n\n\n

Such an idea is currently being put into practice at the Centre for Scientific Computing (CSC) of Goethe University Frankfurt. This is part of the National High Performance Computing (NHR) alliance established in 2021 to bundle the resources and competencies of university high-performance computing \u2013 something like the elite of German supercomputers. The CSC computing facility currently has a performance of a few petaflops \u2013 a petaflop equates to one quadrillion operations per second. Rather than sheer computing power, the fundamental concept is more important. This is known as Modular Supercomputing Architecture (MSA). Thomas Lippert, physicist and computer scientist, who is one of the trailblazers of this concept, wants to build a high-performance computer of a special kind in Frankfurt. He explains what MSA can basically do as follows: \u201cIt enables the parallel processing of computational problems at the system level, we also call this functional parallelism.\u201d But why is it useful? Imagine that a computing task consists of two parts that can only be easily solved on two parallel computers in sequence or at the same time, but which have a very different system architecture.<\/p>\n\n\n\n

<\/div>\n\n\n\n
\"\"<\/figure>
\n

Quite spectacular: in 2021, IBM\u2019s \u201cQuantum System One\u201d at the Fraunhofer Institute for Industrial Engineering in Ehningen was inaugurated as Germany\u2019s first quantum computer. Photo: Andrew Lindemann\/IBM<\/p>\n<\/div><\/div>\n\n\n\n

<\/div>\n\n\n\n

Mixing and matching computer architectures<\/h3>\n\n\n\n

A CPU cluster and a GPU cluster, for example, have such different architectures. CPUs, central processing units, have a limited number of cores and are less suitable for highly parallel data processing.By contrast, GPUs, or graphics processing units, consist of thousands of very small cores that work on a problem at the same time. Two fundamentally different ways of working, but the modular architecture brings them together. \u201cIt is a simple concept and provides the necessary tools and software environments so that both systems can be used at the same time,\u201d explains Lippert. \u201cThis optimizes the entire process in terms of time and energy consumption.\u201d How this works in practice can be seen from computer simulations for the European Space Agency\u2019s space weather observatory that model solar winds \u2013 charged particles (plasma) ejected in solar flares \u2013 and how they interact with Earth\u2019s magnetosphere. The simulation system\u2019s workflow requires two solvers. The field solver calculates the development of Earth\u2019s electromagnetic field \u2013 this part is done by a CPU cluster. The particle solver calculates the motion of the charged particles in the electromagnetic field calculated by the field solver that runs on a GPU cluster.<\/p>\n\n\n\n

Back in Frankfurt, the CSC plans to divide computing tasks in a similar way: a first pilot quantum computer module will be installed this year, teaming up classic high-performance computing and quantum computing in the future. But this cannot be achieved without further ado, as quantum computers work differently to classical computers. The latter process information as bits (binary digits) which are always in exactly one of two possible states: 0 or 1.\u202fQuantum computers, on the other hand, make use of the seemingly bizarre laws of the quantum world, that is, a world with the smallest of particles: photons or quarks. Quantum computers process information in quantum bits or qubits. These can be 0 and 1 at the same time, allowing for an infinite number of possible states in between. The coefficients can even be complex numbers that go beyond real numbers. Qubits are comparable to a rotating coin: it shows not only heads or tails, but all possible \u201cfaces\u201d at once.<\/p>\n\n\n\n

<\/div>\n\n\n\n
\n
\n

IN A NUTSHELL<\/strong><\/p>\n\n\n\n