Skip to main content
PDC Center for High Performance Computing

About the Dardel HPC system

Dardel is an HPE Cray EX supercomputer. This NAISS flagship system was installed at PDC during 2021 and 2022.

Several years ago, the Swedish National Infrastructure for Computing (SNIC) invested in a new high-performance computing (HPC) system for Swedish researchers. The new flagship system is an HPE Cray EX supercomputer, called Dardel, that was installed at PDC during 2021 and 2022. At the start of 2023, the National Academic Infrastructure for Supercomputing in Sweden (NAISS) took over from SNIC, and PDC became a member branch of NAISS.

If you are interested in seeing what is inside the HPE Cray EX supercomputer systems in general, you can explore here .

The Dardel system at PDC consists of two main partitions:

  • a large CPU partition (that is suitable for a broad range of computational applications), and
  • a partition based on graphics processing units (GPUs), which is intended for the most computationally demanding applications.

Installation of the CPU partition and the storage system started in August 2021. This first phase of the system (providing 70,912 CPU cores) has been available for research since December 2021. The GPU partition was installed in late 2022, and has been available to researchers from the spring of 2023.

System

CPU partition

Each compute node has two AMD EPYC™ Zen2 2.25 GHz 64-core processors, which means that each compute node has a total of 128 physical CPU cores. Two virtual hardware threads are enabled on each physical CPU core, which means that each compute node has a total of 256 logical/virtual cores.

There are several memory sizes in the CPU partition.

  • 700 × 256 GB (NAISS thin nodes)
  • 268 × 512 GB (NAISS large nodes)
  • 8 × 1024 GB (NAISS huge nodes)
  • 10 × 2048 GB (NAISS giant nodes)
  • 36 × 256 GB (KTH industry/business research nodes)
  • 248 × 512 GB (KTH industry/business research nodes)

This gives a total of 1270 nodes.

GPU partition

The GPU partition comprises 56 GPU nodes, each of which is configured with

  • one AMD EPYC™ processor with 64 cores (which is a special version of a 64-core AMD CPU that is known as the 7A53 (Trento) and only available from HPE),
  • 512 GB of shared fast HBM2E memory (which is cache-coherent and hence simplifies programming), and
  • four AMD Instinct™ MI250X GPUs (with an impressive performance of up to 95.7 TFLOPS in double precision when using special matrix operations) that are connected by AMD’s Infinity Fabric® as shown in the figure below.
Dardel GPU node architecture

Interconnect

  • HPE Slingshot (ethernet-based) using Dragonfly topology
  • 200 Gb/s since March 2023
High-speed network Slingshot topology as used in the Dardel system

Performance

The LINPACK performance of the first phase of the CPU partition is 2.28 Petaflops and for the GPU partition, it is 8.26 Petaflops.

Usage

  • available for academic research via NAISS
  • available for business and industrial research projects
  • operation currently guaranteed to be funded till Q3 2026