Penguin Solutions has announced upgrades to its OriginAI infrastructure portfolio, adding support for the NVIDIA DGX B300 and RTX PRO 6000 graphics processing units (GPUs). The new configurations, scheduled for general availability in early 2026, aim to extend high-density AI training and inference performance for data center-scale deployments.
According to Penguin Solutions, the NVIDIA DGX B300, powered by the Blackwell architecture, delivers up to 72 petaFLOPs of training throughput and 144 petaFLOPs of inference performance. The RTX PRO 6000 GPU, offered as a dual-slot PCI Express Gen 5 card with 96 GB of GDDR7 memory, is designed to minimize data center space and power requirements while scaling AI deployments cost-effectively.
The OriginAI platform provides scalable reference architectures starting from quarter-pod configurations up to 90 pods, supporting clusters from 64 to more than 24,000 GPUs. The system integrates GPUs from NVIDIA, along with storage options from DDN, Pure Storage, VAST Data, and VDURA, to create a high-performance, pre-validated AI infrastructure. Penguin Solutions claims that its in-factory burn-in and integration environment ensures cluster performance and production readiness before shipment.
A key component of OriginAI is the ICE ClusterWare software suite. The latest release, announced at SC25 in St. Louis, introduces multi-tenancy features for complex multi-user environments, as well as anomaly detection and auto-remediation, which identify and resolve issues such as silent component degradation that may impact cluster performance.
In the press release, Dara Ambrose, vice president, products and solutions for Penguin Solutions, stated, “Integrating the latest NVIDIA GPUs into our OriginAI infrastructure solution is an important development in our ability to help organizations operationalize their AI investments,” and added, “OriginAI provides high-performance solutions that enable scalable AI workloads within dynamic technological environments. By integrating advanced hardware, intelligent management software, and expert services, Penguin Solutions delivers a robust, end-to-end infrastructure solution for the entire AI project lifecycle—from training through inference—helping organizations maintain a competitive edge.”
Penguin Solutions highlights its role as an NVIDIA DGX-Ready Managed Services Provider, combining the new GPUs with validated infrastructure and managed services to support critical data center workloads in AI and high-performance computing environments.
Source: Penguin Solutions
