NVIDIA DGX-1 Clusters

If your workflow needs more compute performance, multiple DGX-1 supercomputers can be linked together in a cluster. Deep learning and AI analysis can be straining for technology and requires vast amounts of power for your network. A cluster will be your bridge to the new avenues that you've been searching for. Prepare to take your network to the next level.

A GPU Cluster isn't as simple as boosting an application with multiple powerful GPUs. There are three key components that make up a GPU cluster: the host nodes, its GPUs and interconnects. With the GPUs powering the vast majority of the calculations, the host nodes and the network interconnect performance' need to match the GPUs to power an even-powered system. Full utilisation is enabled by matching the host memory with the amount of memory on the GPUs, simplifying the process that boost the full development of your applications.

Multi-node DGX-1 small cluster

Supports up to 12 DGX-1 Nodes

Multi-node NVIDIA DGX-1 small cluster  - Novatech Deep Learning

Assume growth for up to 12 nodes

2 racks, 2 IB switches (36 ports)

19.2 kW per rack, can be split across the rack if required

Full bi-section bandwidth for each group of 6 nodes

2:1 oversubscription between groups of 6

Multi-node DGX-1 MEDIUM cluster

Supports up to 36 DGX-1 Nodes

Multi-node NVIDIA DGX-1 medium cluster  - Novatech Deep Learning

Defines a DGX-1 "POD"

Can be replicated for greater scales, eg. Large cluster configuration

6 racks, 6 nodes per rack

Larger IB director switch (216 ports) with capacity for more PODS via an unused port

Multi-node DGX-1 Large cluster

Supports up to 144 DGX-1 Nodes (4 "PODS")

Multi-node NVIDIA DGX-1 large cluster  - Novatech Deep Learning

Implements 4 DGX-1 PODs

Distributed across 24 racks

Full bi-section bandwidth within POD, 2:1 between PODs

Training jobs ideally scheduled with a POD to minimise inter-POD traffic

Learn what a DGX-1 cluster could do for you. Enquire now.