Get access to experimental systems

Researchers and IT experts from SURF member institutions and other Dutch research organizations can get access to experimental systems that we are evaluating in the context of the SURF Open Innovation Lab. This makes it easy for you to adapt your codes, perform benchmarks and follow a co-design approach.

Vrouw met laptop naast glazen serverwand

Projects on experimental systems

Projects that are currently running or have been completed on the experimental systems:

Binding of Covid-19 inhibitors explained by quantum chemical calculations

COVID-19 spike protein

Ana Da Cunha - SURF, Remco W.A Havenith - University of Groningen

Scientific Area: Computational Chemistry
System: Kleurplaat (AMD)

Finding inhibitors for Covid-19 is, given the current global situation, of utmost importance. The extended study of Smith et al. has directed us to potential inhibitors for Covid-19. However, for the development of an efficient design strategy for potent inhibitors, mechanistic details about the bonding of the ligand in the active site of the virus have to be known. In this project, we perform density functional theory and quantum chemical calculations on several potential inhibitors, selected from the study of Smith et al, to explain how the molecule binds to the active site. The findings of this study will be valuable in the search for new Covid-19 inhibitors.

Status: completed


Remo Havenith - University of Groningen

Scientific Area: Computational Chemistry
Systems: Kleurplaat (AMD) & Spider-GPU

Benchmarking of GAMESS-UK in general on both machines (DFT/MP2/hessian/MASSCF/CI) codes and in particular the Valence Bond module TURTLE. All parts have been parallelised. Comparison of running on the AMD/Intel compared to Cartesius will be made. Parts of TURTLE have just been ported to the GP, using openACC (work to rewrite it in CUDA is in progress). The CUDA implementation will also be tested on the AMD and NVIDIA GPUs and compared to performance on Cartesius.

Status: ongoing

Analyzing and optimizing energy efficiency of GPU applications

Ben Werkhoven - Netherlands eScience Centre

Scientific Area: Computer science
Systems: Kleurplaat (AMD) & Spider-GPU

Description : Applications in industry and many scientific fields are enabled and accelerated by Graphics Processing Units (GPUs), at a huge energy cost. While GPUs are relatively energy efficient processors, energy consumption depends on how well the application is optimized to efficiently use the underlying hardware. The optimization of GPU applications is a complex problem that requires finding the best performing combination of many implementation choices and parameters in a very large and discontinuous search space. As such, auto-tuning, the process of a! utomatically searching for the best performing kernel configuration, is often used to optimize the performance of these applications. Despite the great potential of software improvements to the energy efficiency of compute-intensive codes, research in auto-tuning for energy efficiency still remains in its infancy. In this project, we will experiment with new methods for auto-tuning the energy efficiency of GPU applications and benchmark the effectiveness of the these methods on different hardware.

Status: ongoing


Prof. dr. Matthias Moller, dr. Nauman Ahmed - TU Delft, Applied Mathematics Department

Description: see TU Delft website

Status: ongoing


Francisco da Silva, David Maier - QuTech

Description: Running the memory bounded simulations (states sampling) for the QIA project.

Status: ongoing

Emmulation of Variational Quantum Eigensolver (VQE)

Joris Kattemölle - UvA/CWI

Description: The goal of a VQE is to find (or approximate) the ground state energy of the system that is handed to it. VQEs are intended to run on a classical computer which uses a quantum computer in a subroutine. There are many open questions on how these VQEs can be best implemented. Being short of real quantum computers that perform this task (at least for the next couple of years), they want to emulate this entire process on a classical cluster, the Lisa cluster, and find out how a specific VQE can best be implemented. 

Status: ongoing

Performance QAQO

dr. Florian Speelman, Jordi Weggemans - UvA/CWI

Description: Studying the performance of several variants of the QAOA algorithm for a certain graph clustering problem.  

Status: ongoing


dr Eleanor Cerri - SURF, LIACS/LION University of Leiden

Description: Quantum computing for quantum chemistry

Status: ongoing

Optimization of cut-offs to boost performance of quantum key distribution tasks

Tim Coopmans - QuTech

Description: optimization of cut-offs for large quantum-repeater networks that have been constructed from currently-available quantum technology. 

Status: ongoing

Calculating surface code thresholds for distributed quantum computation in the presence of realistic noise and decoherence

Sebastian de Bone - QuTech

Description: in the project, we will simulate an experimentally justified imperfect distributed surface code, by calculating its thresholds with respect to one of the experimental parameters. Entanglement will be distributed stochastically and decoherence channels will be applied to the physical qubits. The models for noise and decoherence that will be implemented will be based on realistic (i.e., experimentally justified) parameters from the laboratory. 

Status: ongoing

Trying Deep Learning benchmarks on AMD infrastructure

Scientific Area: Informatics
System: AMD (Kleurplaat)

Description: Comparison of AMD vs Nvidia GPUs or representative deep learning benchmarks. 

Status: ongoing

Researching the feasibility of using AI on-board of small satellites

F. van Veelen - TUDelft, Hyperion Technologies

Scientific area: Earth Observation
System: Spider-GPU

Description: Small satellites usually have a very limited bandwidth to transmit data to earth. Thus it is not possible to transmit all useful payload data, resulting in inefficient missions. On-board use of AI to process and extract information from the payload data drastically reduces the data volume to be transmitted to earth. Thus the data link to earth is used to its full potential, while most of the information is made available to users. The goal is to develop a method of gathering datasets that can be used to train the AI in a transmission-efficient manner. The findings of this study are applicable to a wide range of fields that benefit from satellite data.

Status: ongoing

Simulating the influence on lipid dynamics on protein-protein interactions

Manuel Menlo, ITQB NOVA

System : AMD (Kleurplaat)

Description : The lipidic fingerprint around a membrane protein is known to affect its dynamics and interactions with other proteins. What is not characterized is how protein dynamics, and its influence on lipid dynamics can actively drive protein interactions in the membrane. We will simulate how protein motion affects the immediate solvating lipid shells, and if/how that effect can promote protein interaction and aggregation. We will employ coarse-grain molecular dynamics simulations to tackle the large scale of the required sy! stems, consisting of tens to hundreds of proteins. 

Status : Ongoing 

Testing a FE2 Framework on AMD Servers

Charles Moulinec, STFC Daresbury Laboratory

System : AMD (kleurplaat)

Description : A FE2 framework designed to simulate material deformation by coupling macro- and micro-scale will be tested on AMD GPUs. The software used at the macro-scale is Alya, which belongs to the PRACE UEABS. It is coupled with Micropp for the micro-scale. The macro-scale part will be run on the CPUs (MPI only) and the micro-scale part on the GPUs.

Status : Ongoing