DGX SuperPOD Architecture

The DGX SuperPOD architecture is a combination of DGX systems, Ethernet networking, InfiniBand Networking, management nodes, and storage. Figure 2 the rack layout of a single SU. With DGX SuperPOD with DGX B300 systems, we utilize standard racks and with traditional power supplies and PDUs.

In our reference design, four DGX B300 fit within a single rack. The rack-level power consumption per rack is ~56kW. The rack layout can be adjusted to meet local data center requirements, such as maximum power per rack and rack layout between DGX systems and supporting equipment to meet local needs for power and cooling distribution.

This Reference Architecture is focused on traditional PDU and AC powered EIA racks. DGX SuperPOD with DGX B300 systems is also available for more dense DB Busbar solutions as well.

Figure 2 shows 72 x NVIDIA DGX B300 PS systems in standard racks each with three (3) 2U rack PDUs for maximum redundancy. Note that depending on your data center’s capability, you might need to reduce the number of DGXs hosted on the same rack.

Figure 2 DGX B300 in Racks

Figure 3 shows an example management rack configuration with networking switches, management servers, storage arrays, and UFM appliances. Sizes and quantities will vary depending upon models used. This example is for 1SU.

Figure 3 Management Equipment in Rack

This reference architecture is focused on 8 SU units with 576 DGX nodes. DGX SuperPOD can scale to much larger configurations up to and beyond 72 SU with 2000+ DGX B300 nodes. See Table 3 for more information.

Table 3 Example Compute Fabric Components for DGX SuperPOD Counts

SU Count	Node Count	GPU Count	Cable Count
			Leaf	Spine	Node-Leaf	Leaf-Spine
1	72	576	8	4	576	576
2	144	1152	16	8	1152	1152
4	288	2304	32	18	2304	2304
8	576	4608	64	36	4608	4608
18	1296	9216	144	72	10368	10368

Contact NVIDIA for information regarding DGX SuperPOD solutions of four scalable units or more.