Dardel
Dardel is a general use HPE Cray EX supercomputer. It combines both a CPU partition so that it can support a broad range of HPC applications and a GPU partition so that it can also support the most computationally demanding applications. It also has a wide variety of memory configurations on the CPU nodes (up to 2TB on some nodes) to support demanding applications.
Do you want to request support from EXCELLERAT staff or consulting on this HPC Resource?
*Registration is required to send inquiries.
HPC Infrastructure Description
Dardel is a HPE Cray EX supercomputer with both CPU and GPU partitions. The CPU partition consists of 1270 nodes each with dual AMD EPYC 64 core processors and between 256GB and 2048GB of memory. Dardel's accelerated partition consists of 56 nodes each with 4 AMD Instinct MI250X GPUs. A M250X GPU contains 2 Graphics Compute Dies (GCDs), and so there are 8 GCDs per node. The nodes are connected via HPE Cray slingshot network. Storage is provided by an 18PB Lustre file system. The LINPACK benchmark performance is 4.08 Petaflops for the CPU partition and 8.26 Petaflops for the GPU partition.
Access
Dardel is operated by PDC, which is a member branch of the National Academic Infrastructure for Supercomputing in Sweden (NAISS). Applications to use the machine for academic research can be made via NAISS at supr.naiss.se/round/compute. Dardel is also available for business and industrial research projects and potential users should contact PDC by email at .
More information
For more information, follow this link.
Name of hosting organisation and Location
PDC, KTH, Teknikringen 14, fourth floor 114 28 Stockholm, Sweden
