Appendix A#
NVIDIA Enterprise Reference Architecture: NVIDIA L40S and NVIDIA Spectrum Platforms#
The NVIDIA L40S and NVIDIA Spectrum Platforms Enterprise RA is optimized for multi-node AI or hybrid visual computing applications. This modular architecture is based on NVIDIA-Certified OVX L40S systems, each equipped with up to four L40S GPUs. Using a four-node scalable unit (SU), this can scale up to 32 NVIDIA-Certified OVX L40S systems, totaling 128 L40S GPUs. Fully tested systems can scale to twenty-four SUs, with the potential for larger clusters based on customer requirements. The flexible rail-optimized end-of-row network architecture accommodates modifications in rack layout and the number of servers per rack. Hardware support is provided through the fulfillment system partner, while software support from NVIDIA is available via a per GPU paid subscription to NVIDIA AI Enterprise.
Use Cases#
Visual Computing: 3D Graphics, Rendering
AI Inference: Medium model parameter inference workloads
AI Training: Small model training and fine-tuning
NVIDIA OVX L40S Reference Configurations#
This Enterprise RA leverages NVIDIA OVX L40S systems, which deliver powerful AI and visual computing performance to accelerate the next generation of AI-enabled enterprise workloads in the data center. The building blocks of the OVX architecture are the performance-optimized NVIDIA L40S GPU server configurations with BlueField-3 DPUs and BlueField-3 SuperNICs. The NVIDIA L40S GPU, based on the Ada Lovelace architecture, is the most powerful universal GPU for the data center, delivering breakthrough multi-workload acceleration for AI, graphics, and video applications. With accelerated AI compute and best-in-class visual computing capabilities, the L40S GPU provides end-to-end performance to power the next generation of AI-enabled data center workloads. NVIDIA-Certified OVX L40S systems are based on a common system design with flexibility for optimizing the configuration to match cluster requirements. Systems are available in 4-GPU and 8-GPU configurations. This was built using the 4-GPU pattern (2-4-3-200 (CPU-GPU-NIC-Bandwidth)), but the 8-GPU pattern (2-8-5-200 CPU-GPU-NIC-Bandwidth) can also be utilized based on specific needs.
Figure 6. 4-GPU OVX system configuration
