SuperMicro Computer, Inc., a leading provider of total IT solutions in AI, cloud, HPC, storage, 5G/Edge and green computing, announced the addition of new products to Supermicro’s plug-and-play AI infrastructure solution, the SuperCluster portfolio, for the enterprise-scale NVIDIA Omniverse™ platform that enhances 3D graphics workflows with high-performance AI. The new SuperCluster is powered by the latest Supermicro NVIDIA OVX™ system and can easily scale as workloads grow.
“Supermicro has led the industry in developing GPU-optimized products for traditional 3D graphics and application acceleration, and now we are expanding into AI as well,” said Charles Liang, president and CEO of Supermicro. “With the rise of AI, businesses are looking for computing infrastructure that brings all these capabilities together in one package. Supermicro’s SuperCluster is ready for NVIDIA Omniverse with fully interconnected 4U PCIe GPU NVIDIA-Certified Systems™ that can house up to 256 NVIDIA L40S PCIe GPUs in each scalable unit. The system can integrate generative AI to deliver high performance across the Omniverse platform. By developing SuperCluster for Omniverse, we are not just delivering a product, we are paving the way for the future of application development and innovation.”
SuperCluster for NVIDIA Omniverse expands Supermicro’s range of application-optimized AI rack solutions. Professionals ranging from product design to industrial digital twins require compute-intensive 3D graphics workflows. Generative AI extends existing 3D graphics workflows and accelerates a new era of applications. SuperCluster for NVIDIA Omniverse™ simplifies the deployment of scale-out infrastructure to meet the needs of multi-workload 3D graphics and AI.
Supermicro NVIDIA OVX systems are the foundational building blocks for the compute power of the cluster. Each system node features up to eight of the latest NVIDIA PCIe GPUs, combining the best 3D graphics and virtual reality performance with Tensor Cores and Transformer Engine support for outstanding generative AI performance. The system includes 4x 2700W Titanium Level redundant power supplies, all housed in a superior airflow enclosure to ensure stability in high utilization scenarios. Up to four NVIDIA BlueField®-3 SuperNICs or four NVIDIA ConnectX®-7 NICs per system provide multi-port 400Gb/s networking with high scalability and security.
Supermicro’s 4U PCIe GPU systems have passed a rigorous validation process testing performance, reliability, scalability and security and are certified by NVIDIA for use in NVIDIA Omniverse™ to maximize performance for diverse workloads within the NVIDIA Omniverse development platform, including the OpenUSD ecosystem and generative AI techniques via Omniverse Cloud APIs.
Also Read: Lambda Launches First Self-Serve, On-Demand NVIDIA HGX H100 and NVIDIA Quantum-2 InfiniBand Clusters for AI Model Training
NVIDIA Omniverse™ SuperCluster is a fully interconnected infrastructure solution that gives developers, engineers and others access to the highest levels of GPU computing when they need it with seamless access to virtual GPUs or bare metal access to full system nodes. The 400Gb/s high-performance network fabric provided by the NVIDIA Spectrum™-X Ethernet platform enables companies developing custom large language models to tap into a unified memory pool of GPU memory across system nodes, which is essential for training large-scale AI models.
Supermicro’s validated rack solutions range from 4 GPUs to 256 GPU scalable units and can scale to suit any size business. Customers can get started right from day one with plug-and-play racks that are thoroughly validated at the L12 level.
Highly customizable solutions from single rack to enterprise scale
NVIDIA Omniverse™ SuperCluster offers a wide range of sizes and options depending on your requirements. System nodes can have 4 or 8 GPUs per system, and deployment sizes range from a single rack with 4 systems to a scalable unit with 32 systems across 5 racks. Larger deployments can be further expanded via scalable units to create clusters of virtually any size.
SuperCluster (scalable unit) for NVIDIA Omniverse ™ includes:
- 32 Supermicro SYS-421GE-TNRT (dual root) or SYS-421GE-TNRT3 (direct attach) PCIe GPU system nodes
- 256 or 128 NVIDIA L40S GPUs
- 3 Supermicro SYS-121H-TNR Hyper System Control Nodes
- 3 x 400G 64-port NVIDIA Spectrum™ SN5600 Ethernet Compute Fabric Switches
- 2 x 400G 64-port NVIDIA Spectrum SN5600 Ethernet storage/control fabric switches
- 2 x 1G 48-port NVIDIA Spectrum SN2201 Ethernet Managed Switches
- NN NVIDIA BlueField-3 SuperNICs or NVIDIA ConnectX-7 NICs
- 5 Rack: 48U 750mmx1200mm
A SuperCluster for NVIDIA Omniverse ™ can be configured starting from a single rack , which includes :
- Four Supermicro SYS-421GE-TNRT or SYS-421GE-TNRT3 PCIe GPU system nodes
- 16 or 8 NVIDIA L40S GPUs
- 2 Supermicro SYS-121H-TNR Hyper System control nodes
- One 400G 64-port NVIDIA Spectrum SN5600 Ethernet Compute Fabric Switch
- 1 x 400G 64-port NVIDIA Spectrum SN5600 Ethernet Storage/Control Fabric Switch
- 1 x 1G 48-port NVIDIA Spectrum SN2201 Ethernet Managed Switch
- NN NVIDIA BlueField-3 SuperNIC or NVIDIA ConnectX-7 NIC
- 1 rack: 48U 750mmx1200mm
Source: PRNewswire