#USA #Supermicro #San Jose #AI Storage #Grace CPU

Supermicro Unveils a Groundbreaking Petascale All-Flash Storage Server

At the GTC 2025 Conference, Supermicro, Inc. (NASDAQ: SMCI), a leading provider of comprehensive IT solutions for AI/ML, cloud computing, and 5G/Edge, has showcased a remarkable new all-flash storage server. This innovative system is designed to handle demanding software-defined storage workloads, emphasizing Supermicro's commitment to cutting-edge technology and efficiency in computing.

The company has emphasized that its latest storage server leverages the NVIDIA Grace CPU Superchip, which features 144 Arm Neoverse V2 cores. This powerful architecture is specifically tailored for high-performance workloads typical in AI and machine learning applications. Supermicro has crafted this storage solution to maximize bandwidth and performance, fully utilizing the capabilities offered by PCIe Gen5 SSDs with linear scalability to keep pace with modern processing demands.

Collaboration with NVIDIA and WEKA

In a strategic partnership, Supermicro collaborated closely with NVIDIA and WEKA to ensure this solution not only delivers high performance but also emphasizes energy efficiency. The Supermicro Petascale storage server can support up to 16 hot-swap EDSFF PCIe Gen5 E3.S NVMe drives, allowing for impressive total raw capacity of 983TB when fully utilized. Furthermore, a rack configuration containing 40 units can potentially offer up to 39.3PB of storage, making it an ideal option for enterprises with substantial data retention needs.

Charles Liang, the President and CEO of Supermicro, commented on the innovation stating that, "The Supermicro Petascale server underlines our dedication to innovation and customer choice. We have developed this server with the goal of enabling advanced AI workflows without compromise, ensuring it stands out in today’s competitive market."

Designed for Scalability and Efficiency

The ARS-121L-NE316R 1U storage system demonstrates Supermicro's Building Block Solutions® design philosophy, which encompasses symmetrical architecture optimized for scaling AI workloads. This architecture not only simplifies server management but also enhances overall efficiency. Given the increasing importance of sustainability, the energy-efficient nature of the NVIDIA Grace CPU Superchip stands to reduce power consumption while maintaining high performance levels.

Rob Davis, NVIDIA's Vice President of Storage Networking Technology, noted that this collaboration has provided storage software vendors with flexible hardware choices to meet diverse application requirements. The enhancements, particularly in network protocol acceleration through NVIDIA’s GPUDirect Storage, present a significant advantage over traditional x86-based servers.

Partnership with WEKA for High Performance

Supermicro is also proud to work alongside WEKA, famed for its robust high-performance storage software tailored for AI, HPC, and data management tasks. This cooperative effort emphasizes a unified strategy to develop an architecture that promotes significant performance and energy efficiency. Nilesh Patel, Chief Product Officer at WEKA, shared excitement regarding the partnership and highlighted remarkable results from initial testing of their software on Supermicro’s Petascale storage server, showcasing notable performance scaling and energy savings.

Unveiling at GTC 2025

Visitors at GTC 2025 can look forward to an in-depth presentation of the new Petascale storage server during on-demand sessions, where Supermicro will delve deeper into the operational capabilities and benefits that this advanced server offers.

To dive into the various features of Supermicro's innovative storage solutions, you can visit their official website. As they continue to enhance their product offerings, Supermicro remains at the forefront of IT solutions designed for a tech-driven future that prioritizes both performance and sustainability in an increasingly digitized world.