I'm glad to announce you the MetaCentrum's computing capacity was extended with new clusters:
zia.cerit-sc.cz (location Brno, owner CERIT-SC), 5 nodes, 640 CPU cores, GPU card NVIDIA A100, in each node:
The cluster can be accessed via the conventional job submission through PBS batch system (@pbs-cerit server) in gpu priority and short default queues
The cluster is equipped with currently the most powerful graphics accelerators NVIDIA A100 Tensor Core GPU (https://www.nvidia.com/en-us/data-center/a100/). It delivers unprecedented acceleration at every scale to power the world’s highest-performing elastic data centers for AI, data analytics, and HPC.
The main advantages of the NVIDIA A100 include a specialized Tensor core for machine learning applications or large memory (40 GB per accelerator). It supports calculations using tensor cores with different accuracy, in addition to INT4, INT8, BF16, FP16, FP64, a new TF32 format has been added.
On CERIT-SC GPU clusters, it is possible to use Docker images from NVIDIA GPU Cloud (NGC) - the most used environment for the development of machine learning and deep learning applications, HPC applications or visualization accelerated by NVIDIA GPU cards. Deploying these applications is then a matter of copying the link to the appropriate Docker image, running it in the Docker container (in Podman, alternatively in Singularity). More information can be found at https://wiki.metacentrum.cz/wiki/NVidia_deep_learning_frameworks
For complete list of available HW in MetaCentrum see http://metavo.metacentrum.cz/pbsmon2/hardware