Supermicro has introduced a plug-and-play SuperCluster tailored for the NVIDIA Omniverse platform.
This rack solution accommodates up to 256 cutting-edge NVIDIA PCIe GPUs within a scalable unit meticulously optimised for large-scale deployments. Designed to enhance performance for 3D and AI workloads, the SuperCluster streamlines the deployment of scale-out infrastructure to meet diverse workload demands.
Equipped with Supermicro NVIDIA OVX systems, it offers enterprises a flexible, resource-optimised solution for generative AI-enhanced 3D workflows on an enterprise scale.
The primary goal is to empower high-performance, AI-driven 3D workflows at an enterprise level. By using Supermicro’s NVIDIA OVX systems, businesses gain scalability as their workloads expand.
Supermicro has long been involved in GPU-optimised product development, initially focusing on 3D graphics and application acceleration. Now, the company has shifted its focus to AI.
Supermicro CEO Charles Liang said: “As AI gains prominence, enterprises seek unified computing infrastructure that encompasses all these capabilities. Supermicro’s SuperCluster features fully interconnected 4U PCIe GPU NVIDIA-Certified Systems for NVIDIA Omniverse, supporting up to 256 NVIDIA L40S PCIe GPUs per scalable unit. This system delivers exceptional performance across the Omniverse platform, including generative AI integrations.”
The SuperCluster for NVIDIA Omniverse streamlines the deployment of scale-out infrastructure to meet the multi-workload demands of 3D and AI. Supermicro NVIDIA OVX systems serve as the fundamental building blocks of the cluster’s computing power.
Each system node accommodates up to 8 of the latest NVIDIA PCIe GPUs, providing a powerful combination of 3D graphics, virtual reality performan ce, and exceptional generative AI capabilities via Tensor Cores and Transformer Engine support.
SuperCluster for NVIDIA Omniverse offers interconnected infrastructure, ensuring seamless access to high-level GPU computing for developers, artists, and engineers. Whether through virtual GPUs or direct access to full system nodes, Supermicro enables efficient scaling.
The inclusion of the NVIDIA Spectrum -X Ethernet platform, delivering a 400Gb/s high-performance network fabric, facilitates GPU memory pooling across system nodes—an essential feature for training large AI models.
