Euler VIII is operational

From ScientificComputing
Jump to: navigation, search


The new Euler VIII nodes are now available to the cluster users. In the initial phase, the nodes can run short jobs from the 4 hour and the 24 hour queue and will be added to the longer queues later. They have

  • Two 64-core AMD EPYC 7742 processors (the same as Euler VI) which results in a total of 128 cores
  • 512 GB of DDR4 memory clocked at 3200 MHz
  • 920 GB of local scratch space

Select or avoid Euler VIII nodes

Euler VIII nodes are connected to the cluster's 100G Ethernet network, which has a higher latency than the cluster's InfiniBand (IB) network. If you are running jobs that require a low-latency communication, then please avoid running on Euler VIII and request nodes connected to the IB network.

To force your job to run on Euler VIII, request the -C fabric8 sbatch option:

sbatch -C fabric8 --wrap="./my_command"

To prevent your job from running on these nodes, request the -C ib sbatch option:

sbatch -C ib --wrap="./my_command"

Known issues

  • Nothing so far


If you encounter any problem with running your jobs on the new Euler VIII nodes, then please report it to cluster support. Please note that during the Christmas holidays, when ETH is closed, there will be no cluster support. Problems reported during this time will be taken care of after the Christmas holidays.