Introducing Supermicro’s New Supercluster
Supermicro, Inc., a leading IT solution provider specializing in AI, cloud, storage, and 5G/edge solutions, has announced a revolutionary addition to its Supercluster portfolio. The new Supercluster is designed specifically for the NVIDIA Omniverse platform, offering high-performance generative AI-enhanced 3D workflows at an enterprise scale. This system features the latest Supermicro NVIDIA OVX systems, ensuring seamless scalability as workloads expand.
A Leader in GPU-Optimized Products
“Supermicro has led the industry in developing GPU-optimized products, traditionally for 3D graphics and application acceleration, and now for AI,” said Charles Liang, President and CEO of Supermicro. With the rise of AI, enterprises require computing infrastructure that combines these capabilities into a single package. The new Supercluster integrates up to 256 NVIDIA L40s PCIe GPUs per scalable unit, offering unparalleled performance across the Omniverse platform, including generative AI integrations.
Advanced Features for Diverse Workloads
The Supermicro Supercluster broadens the scope of application-optimized AI rack solutions. The Supercluster is built with compute-intensive 3D workflows in mind, catering to a wide range of professionals from product designers to industrial digital twins. Each system node hosts up to eight of the latest NVIDIA PCIe GPUs, delivering top-notch 3D graphics and virtual reality performance, alongside exceptional generative AI capabilities through tensor cores and transformer engine support.
Unmatched Scalability and Flexibility
One of the standout features of the Supercluster for NVIDIA Omniverse is its high scalability. Systems can be scaled from a single rack with four systems to a scalable unit with 32 systems in five racks. Further scalability can be achieved through the addition of scalable units, fitting enterprises of virtually any size. The system nodes are powered by 4x 2700W titanium level PSUs within a high-airflow chassis, ensuring stability under high-utilization scenarios.
Comprehensive Validation and Seamless Integration
Supermicro’s 4U PCIe GPU systems are NVIDIA-certified, having passed a rigorous validation process that guarantees performance, reliability, scalability, and security. This ensures that organizations can maximize performance across diverse workloads within the Omniverse platform. The 400 Gb/s high-performance network fabric, provided by the NVIDIA Spectrum-X Ethernet platform, allows enterprises to tap into a combined pool of GPU memory necessary for training large AI models.
Ready-to-Deploy, Customizable Solutions
Supermicro’s validated rack solutions range from systems featuring four GPUs to 256 GPU scalable units, ready for use on day one. These plug-and-play solutions are highly customizable, capable of being tailored to meet customer-specific requirements, ranging from a single rack to an enterprise scale.
