Information on new resources in Kebnekaise 2024-08-26

  • Posted on: 26 August 2024
  • By: ake

Dear users of HPC2N,

we want to inform you about the latest sets of new hardware available on Kebnekaise.

During the spring and summer we have installed some new nodes, both pure
CPU nodes and GPU nodes.

All nodes are based on AMD Zen4 CPUs. The pure CPU nodes have 256 cores/node. 
Most of the GPU nodes have dual Nvidia L40s cards, two nodes have
Nvidia quad-H100 SXM5 cards, and two nodes have 6xL40s cards.

To be able to utilize these nodes we strongly recommend reading the
"The Batch System" section in our updated documentation on https://docs.hpc2n.umu.se/

The best way of using GPU's is to no longer specify exact card to use but instead ask for the features needed from the GPU card, for instance double or single precision capability. How to do this is described in the "The different parts of the batch system" subsection of
"The Batch System".
Specifying features instead of a specific card will reduce the job's wait time before it can be started.

Please note the 'Squeue “Reason” explained' section, reading that will reduce confusion when looking at the output of "squeue".

Specifically we currently have (including the old systems)
 - 48 nodes with 2x14 cores Intel Skylake CPUs, 185 GB of usable memory
 - 10 nodes with 2x14 cores Intel Skylake CPUs, 185 GB of usable memory, 2xV100 GPUs
 - 8 nodes with 4x18 cores Intel Broadwell CPUs, 2.8 TB of usable memory

New nodes from last year and this spring
 - 1 node with 2x64 cores Zen3 CPUs, 1000 GB of usable memory
 - 8 nodes with 2x128 cores Zen4 CPUs, 629 GB of usable memory

 - 2 nodes with 2x24 cores Zen3 CPUs, 496 GB of usable memory, 2xA100 GPUs
 - 1 node with 2x24 cores Zen3 CPUs, 496 GB of usable memory, 2xMI100 AMD GPUs
 - 1 node with 2x24 cores Zen4 CPUs, 310 GB of usable memory, 2xA6000 GPUs
 - 10 nodes with 2x24 cores Zen4 CPUs, 310 GB of usable memory, 2xL40s GPUs
 - 2 nodes with 2x48 cores Zen4 CPUs, 620 GB of usable memory, 4xH100 SXM5 GPUs

Nodes that will be taken into production the coming week
 - 1 node with 2x64 cores Zen4 CPUs, ~740 GB of usable memory, 8xA40 GPUs
 - 2 nodes with 2x64 cores Zen4 CPUS, ~740 GB of usable memory, 6xL40s GPUs

The old Broadwell nodes, including the old K80 GPU nodes, were taken out of production last year.

If you have problems with missing software on the new nodes, please inform us at support@hpc2n.umu.se as soon as possible so we can remedy the problem.

Happy computing,
HPC2N staff
 

Updated: 2024-10-10, 12:39