Publication date: 11th October 2022
Large computational infrastructures in academic laboratories and research institutes rely on server racks to fulfill heavy computational duties. They display few disadvantages for research in small institutions and businesses:
(1) Server racks are usually very expensive and difficult to assemble and maintain, requiring large budgets and dedicated personnel.
(2) Server racks require powerful airflow to cool components down. This operation is usually performed by Delta fans (or similar), ventilators pushing remarkably great air flow (252.85 CFM for a 120x120x38 mm PFB1224UHEP0 fan) and static pressure (35.877 mmH2O for the same fan), but also generating a considerable amount of noise (up to 66.5 dB(A) for the same fan). These characteristics do not make server racks particularly appealing for office environments in small business.
(3) The clocking speeds of server CPUs are usually lower (~2-3 GHz) compared to modern workstations (> 4 GHz).
(4) CPU use is generally suboptimal due to the scarce attention of the operator to the proper level of parallelization of the task.
NVIDIA® recently launched on the market a new high-end desktop (HEDT) for intensive computation, machine learning and artificial intelligence: DGX Station A100.1 The workstation is powered by an AMD Epyc 7742 (64 cores) and four NVIDIA A100 cards, the most advanced GPUs based on Ampere technology. MedeA® also launched an assembled in-house workstation called MedeA Instrument2 powered by one (64 cores) or two AMD Epyc 7000 series CPUs (128 cores). Following the same line of thinking, our research institution, MagnetoCat SL,3 mounted its own HEDTs based on AMD Ryzen Threadripper CPUs for dedicated computational work. These CPUs are fast-clocking multicore (16, 24, 32 and 64 cores) nodes with large cache memory and the ability to use 4-to-8 memory channels of buffered ECC-protected server RAM.
TOC figure shows the computational performances of our workstations in the AUSURF112 benchmark4 for Quantum Espresso v.6.8.5 Top tier performers in this benchmark are: 2xAMD EPYC 75F3 (64 cores per motherboard) with 257±23 s, AMD Threadripper 3970x (32 cores) with 303±3 s and AMD EPYC 75F3 (32 cores) with 304±4 s.
Our first workstation, mounting an AMD Ryzen Threadripper PRO 3995wx (64 cores), 256GB RAM and a NVIDIA Quadro GV100 32GB (Volta), is able to run the CPU-only benchmark in 344 s using MPI-only parallelization. The value can be substantially lowered to a record-holder 182 s when a mixed parallelization of MPI and OpenMP is cleverly implemented. Moreover, 1 core of 3995wx + Quadro GV100 GPU run the benchmark in 69 s (~2.6 times faster than fully optimized benchmark on 64 cores). Our second workstation, mounting an AMD Ryzen Threadripper PRO 3955wx (16 cores), 256GB RAM and a NVIDIA A100 40GB (Ampere), is able to run the benchmark in 48 s using 1 core + A100 GPU, confirming the top performances of NVIDIA A100 as engine for high computation and AI on the market (our institution managed the non-trivial action to implement the A100 in a workstation setup). Thanks to the large PCIe bandwidth, the workstations can also be upgraded to multi-GPU NVLINK systems. From the point of view of chemical accuracy, the action of offloading most of the expensive routines of Quantum Espresso to GPU allows the use of improved cut-offs for plane-waves, finer k-point sampling of the Brillouin zone and more accurate pseudopotentials with smaller potential cores.
Liquid cooling guarantees silent and efficient operations: CPU temperatures are around ~60-65°C at full load (room temperature ~20°C). GPU temperatures max out at around ~50-60°C during full load.
[1] https://www.nvidia.com/content/dam/en-zz/Solutions/Data-Center/dgx-station/nvidia-dgx-station-a100-datasheet.pdf; https://images.nvidia.com/aem-dam/Solutions/Data-Center/nvidia-dgx-station-a100-infographic.pdf
[2] https://www.materialsdesign.com/medea-instrument
[3] www.magnetocat.com
[4] https://openbenchmarking.org/
[5] https://www.quantum-espresso.org/
MF would like to thank SpinCat project and European Union's Horizon 2020 research and innovation program under Grant Agreement no. 964972.
nanoGe is a prestigious brand of successful science conferences that are developed along the year in different areas of the world since 2009. Our worldwide conferences cover cutting-edge materials topics like perovskite solar cells, photovoltaics, optoelectronics, solar fuel conversion, surface science, catalysis and two-dimensional materials, among many others.
Previously nanoGe Spring Meeting (NSM) and nanoGe Fall Meeting (NFM), MATSUS is a multiple symposia conference focused on a broad set of topics of advanced materials preparation, their fundamental properties, and their applications, in fields such as renewable energy, photovoltaics, lighting, semiconductor quantum dots, 2-D materials synthesis, charge carriers dynamics, microscopy and spectroscopy semiconductors fundamentals, etc.
International Conference on Hybrid and Organic Photovoltaics
International Conference on Hybrid and Organic Photovoltaics (HOPV) is celebrated yearly in May. The main topics are the development, function and modeling of materials and devices for hybrid and organic solar cells. The field is now dominated by perovskite solar cells but also other hybrid technologies, as organic solar cells, quantum dot solar cells, and dye-sensitized solar cells and their integration into devices for photoelectrochemical solar fuel production.
Asia-Pacific International Conference on Perovskite, Organic Photovoltaics and Optoelectronics
The main topics of the Asia-Pacific International Conference on Perovskite, Organic Photovoltaics and Optoelectronics (IPEROP) are discussed every year in Asia-Pacific for gathering the recent advances in the fields of material preparation, modeling and fabrication of perovskite and hybrid and organic materials. Photovoltaic devices are analyzed from fundamental physics and materials properties to a broad set of applications. The conference also covers the developments of perovskite optoelectronics, including light-emitting diodes, lasers, optical devices, nanophotonics, nonlinear optical properties, colloidal nanostructures, photophysics and light-matter coupling.
The International Conference on Perovskite Thin Film Photovoltaics Perovskite Photonics and Optoelectronics (NIPHO) is the best place to hear the latest developments in perovskite solar cells as well as on recent advances in the fields of perovskite light-emitting diodes, lasers, optical devices, nanophotonics, nonlinear optical properties, colloidal nanostructures, photophysics and light-matter coupling.