The AgResearch eResearch Infrastructure (eRI) Compute cluster consists of the resources in the table below.
Specifications
PARTITION | SLURM CPUS | S:C:T | NODES | NODELIST | Mem/Node |
|---|---|---|---|---|---|
compute | 256 | 2:64:2 | 6 | compute-[0-5] | 1T |
gpu | 96 | 2:24:2 | 1 | gpu-0 | |
hugemem | 256 | 2:64:2 | 2 | hugemem-[0-1] | 4T |
interactive | 8 | 8:1:1 | 3 | interactive-[0-2] | 15G |
vgpu | 32 | 32:1:1 | 4 | vgpu-[0-3] |
S:C:T stands for sockets : cores : threads
SLURM CPUS refers to the number of consumable CPU resources each node offers the cluster and is dependent on the scheduler’s configuration.
S*C*T = SLURM CPUS (2*64*2=256)
CPUs
All CPUs currently in the cluster are AMD EPYC Milan 7713 processors with a base clock of 2.0GHz and max boost of 3.675GHz.
NB: virtualised compute nodes (e.g. in interactive and vgpu partitions) use the same underlying CPUs but may appear differently within the compute node instance.
GPUs
The gpu partition contains a single node with 2x NVIDIA A100 GPGPUs, PCIe 40GB cards
The vgpu partition contains four virtualised compute nodes, each with a single NVIDIA A10 GPGPU, PCIe 24GB cards