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

SURF Open Innovation Lab currently has the following experimental systems that are available to researchers and the academic community:

AMD servers with 8 MI50 GPU cards per server

Name of the system: Kleurplaat
Maintained and operated by SURF Open Innovation Lab, in collaboration with Nikhef and AMD
Node: Dual socket AMD 7742 Rome CPU (128 cores) server with 512 GB RAM and 8 x AMD MI50 (16 GB) GPUcards per server
Number of servers : 6
Setup: standalone servers
Interconnect:  not yet available (Infiniband in the future)

Intel servers with 1 NVIDIA V100 card per server

Name of the system: Spider-GPU
Maintained and operated by SURF
Node: Dual socket Intel Xeon Gold 6226 CPU (24 cores) server with 768 GB RAM and 1 x NVIDIA V100 (32GB) GPU card per server
Number of servers: 2
Setup: integrated with the Spider platform for Data Processing
Interconnect:  CephFS (shared file system)

High memory nodes

Name of the system: SOIL-nodes
Maintained and operated by SURF
Node: 1.5 TB RAM, 2 NVIDIA TitanRTX GPU cards and 2 sockets CPUs with 20cores/socket (Intel Gold Skylake)
Number of servers: 4
Setup: integrated with LISA
Interconnect:  Ethernet

How to apply

All researchers in the Netherlands are invited to apply for access using this form. International collaborations are also welcome, however it is encouraged for the principal investigator to be based in the Netherlands. 

For the AMD servers we are particularly interested in receiving applications that aim to tackle benchmarking, AMD GPU code development/porting, and workflow optimization. Access can be granted for up to a maximum of 5 months.

For the Intel+NVIDIA servers we are particularly interested in receiving applications that aim to investigate data intensive workflows and GPU code development/optimization for (parts of) such workflows.

For the High memory nodes we are particularly interested in receiving applications related to quantum computing and machine learning.

(Please be aware that coding for GPU's is non-trivial. SURF/SOIL provides access to these systems on a shared risk basis without any warranty or guarantees. Support by SURF beyond standard operation of the cluster is very limited and if required should be mentioned in the application.)



Would like to have more information? Please contact us!