Home > AI Solutions > Gen AI > White Papers > Generative AI in the Enterprise with NVIDIA Spectrum-X Networking Platform > System configuration
The following tables list the system configurations and software stack used for the validation efforts in this design:
Component | Configuration 1 | Configuration 2 |
Compute server for model customization | 8 x PowerEdge XE9680 servers | 4 x PowerEdge R760xa servers |
GPUs per server | 8 x NVIDIA H100 SXM GPUs | 4 x NVIDIA L40S PCIe GPUs |
Backend network adapters | SuperNIC: 8 x NVIDIA Bluefield-3 Single Port 400 GbE QSFP112 PCIe FH (B3140H) | SuperNIC: 2 x NVIDIA Bluefield-3 Single Port 400 GbE QSFP112 PCIe FH (B3140H) |
Frontend network adapters[1] | DPU: 2 x NVIDIA Bluefield-3 Single Port 400 GbE QSFP112 PCIe FH (B3140H) | DPU: 1 x NVIDIA Bluefield-3 Single Port 400 GbE QSFP112 PCIe FH (B3140H) |
[1] While NVIDIA BlueField-3 B3220 is recommended for frontend networking, the lab validation used NVIDIA BlueField-3 B3140H due to hardware availability.
Component | Details |
Operating system | Ubuntu 22.04.1 LTS |
Cluster management | NVIDIA Base Command Manager Essentials 10.24.05a |
Slurm cluster | Slurm 23.02.4 |
Kubernetes cluster | Version 1.28 with:
|
AI framework | NVIDIA NeMo Framework v24.05.01 |