[Berzelius-users] Berzelius is now operational

[Henrik Henriksson]

Dear Berzelius users,

Berzelius is now available again.

The cluster now consists of

- 1.5 PB all flash Lustre storage, of which 0.5 PB is new,
- 60 "thin" Nvidia DGX A100 40 GB nodes,
- 34 new "fat" Nvidia DGX A100 80 GB nodes.

We expect the new 80GB "fat" nodes to be available for users to schedule jobs on
in the near future. In the same manner as on our other clusters, the nodes will
be split up into `fat` and `thin` nodes, so users can schedule jobs on the
desired hardware. GPUs with 80GB RAM will be billed higher than the 40GB
variants, as the 80GB variant is almost twice as expensive. More information on
the new scheduling policy will be available shortly.