New research supercomputer online

Swedish research has now gained access to the first 600 nodes of the new supercomputer Tetralith, which will be the most powerful computer in the Nordic region when it comes fully online this autumn. Researchers at LiU also have access to supercomputer Sigma, financed completely by the university.

Data center at National Supercomputer Centre, Campus Valla, Linköping University. Thor Balkhed

In September, the first users started to exploit the computing capacity of the new supercomputer Tetralith, which will achieve a calculation speed of 4 petaflops (4,000,000,000,000,000 floating point operations per second) when fully online. Tetralith is a Swedish national resource, financed in collaboration between the Swedish Research Council (VR, through the Swedish National Infrastructure for Computing, SNIC) and the Swedish universities. In order to use Tetralith, Swedish researchers must apply for resources at SNAC, the Swedish National Allocations Committee, part of SNIC.

Triolith, the previous supercomputer at the National Supercomputer Centre, NSC, will be dismantled to make space for the remaining nodes. When fully constructed, Tetralith will consist of 1,892 nodes, which together offer the calculation power of 60,644 processor cores.

Sigma a smaller supercomputer

In parallel with the procurement of Tetralith, an extra resource at LiU was acquired. This smaller supercomputer, Sigma, has 108 nodes and 3,456 processor cores, and is intended Igor Abrikosov, professor i teoretisk fysikProfessor Igor Abrikosovsolely for LiU researchers, where resources are booked directly at the NSC.

“This is an extremely important resource for several of the major research groups at LiU”, says Matts Karlsson, director of the NSC, whose own research group is a major consumer of supercomputer resources for the simulation of fluid flow and turbulence around, for example, aeroplanes, timber trucks and, not least, inside the human body.

“It’s marvellous that LiU supports us in this way”, Igor Abrikosov, professor of theoretical physics at LiU, confirms.

Igor Abrikosov and his research group are trying to understand the fundamental physical phenomena that determine the properties of materials.

“It currently takes 20 years to develop a new material: we hope to bring that down to 5-6 years. By making it possible for us to study and simulate increasingly complicated systems, the supercomputers allow us to approach reality and increase the accuracy of the calculations. If we can master the fundamental physical methods, the probability that the simulations are successful is significantly higher”, he says.

Still a limited resource

The theories and methods developed by Igor Abrikosov’s group will subsequently form the basis for developing several promising families of new materials. Examples from research carried out at LiU are hard nitrides used for coatings within manufacturing industry, the production of new alloys for fuel cells, and not least the development of the new perovskite-based materials for environmentally sensitive and efficient light-emitting diodes and solar cells.

“We can use Sigma to prepare and test the simulations we subsequently run at full scale in Tetralith”, says Igor Abrikososv.

Photo credit Thor Balkhed“It’s sometimes necessary to carry out pilot runs when applying for research grants, and here access to Sigma will be invaluable”, says Matts Karlsson.

Sigma will replace a previous resource, Gamma, but even with the increased calculation capacity offered by Sigma and Tetralith, needs are growing more rapidly than processor power.

“We still have to set strict priorities: supercomputer capacity is a very limited resource”, says Igor Abrikosov.

Simulations that require extreme computer capacity are being increasingly used within research. Around ten research groups at LiU are interested in running simulations on Sigma, considerably more than was the case for Gamma. An example of a large new research area that requires computer power is artificial intelligence.

Summit most powerful 

Earlier in the year, what is currently the world’s most powerful supercomputer came online, beating by a wide margin the Chinese systems that have dominated in recent years. The Summit supercomputer at the Oak Ridge Laboratory in Tennessee, a research institution owned by the US Department of Energy, has a calculation speed of 200 petaflops – dedicated to AI research.

Show/Hide content

Related content
Show/Hide content

Latest news from LiU
Show/Hide content