Project objectives.

For the period 2017-2022, the LPEM has put together a research group specialising in correlation-based materials theory with a focus on super and thermal-electric conductors.

LPEM conducts research into these topics using various experimental investigative techniques—DC and AC conductivity measurements, measurements of thermoelectric coefficients, magnetic properties, NMR spectroscopy, scanning probe microscopy, etc.

ESPCI needed new resources to provide the necessary power to simulate theories or confirm physical measurements.


The Bechtle and FRA-SYS solution was different from the standard solutions we had looked into. Despite that, the benefits presented to us quickly became tangible as we stuck with a tower solution that is quick and easy to install and also quiet because it is water-cooled.

Luca De’ Medici, Associate Research Director, LPEM/ESPCI


Solution.

After analysing LPEM’s needs, Bechtle suggested an innovative over-clocking solution based on the dual concepts of:

  • Maximum acceleration and complete use of CPU resources
  • A water-cooled tower model which required no special IT infrastructure (machine room, ambient noise, etc.)

 

Expert researchers’ minimal or understandable lack of HPC expertise was also significant as a solution needed to be found that was not overly complicated and which did not go over budget.

 

The tailored solution, configured with FRA-SYS has 18 cores with 4.2 GHz, 128 GB memory and local storage on NVMe and RAID-based hardware.

 

To ensure the solution is scalable for the future, FRA-SYS replaced the eight original 16 GB modules with 4 32 GB modules for no extra cost.

Grafik ESPCI

Calculated electronic compressibility in FeSe—a new super-conductor with a high critical temperature—dependent on the density and intensity of electronic interactions (U).

 

From P.V. Arribi and L. de’ Medici, Phys. Rev. Lett 121, 197001 (2018)

Business benefits.

The lab has procured a HPC solution that the researchers are able to get to grips with. Compared to the LPEM’s previous computing resources, standard models have been accelerated by an average of 20%

by a solution that is whisper-quiet during operations.

Questions?

*Required fields

If you’d like to know more about how we handle your personal data, please read our Privacy Policy.

Message sent
Send another message