Supermicro Unveils NVIDIA Grace CPU Superchip-Based Servers

NVIDIA

At COMPUTEX 2022 Virtual, Supermicro has announced the deployment of the NVIDIA Grace CPU Superchip in a wide range of servers optimized for AI, HPC, data analytics, digital twins, and compute-intensive applications. On forthcoming Supermicro servers, the full NVIDIA software stack, including NVIDIA AI Enterprise, would work flawlessly.

Supermicro servers will leverage the NVIDIA Grace CPU Superchip to enable a greater range of developers and IT administrators to take use of artificial intelligence (AI) technologies as they spread across numerous sectors.

Photo Charles Liang, President and CEO at Supermicro
“Our first-to-market new server systems can quickly address the rapidly expanding AI, HPC, and data analytics markets,” said Charles Liang, President and CEO at Supermicro.

Supermicro servers will be tuned for energy efficiency, which will lower running expenses across the board. Furthermore, applications would not be constrained by transmission speeds thanks to a memory bandwidth of up to 1 TB/second.

A limited number of servers will be available at first, starting with a 2U 2 Node portfolio and gradually expanding the product range.

“Supermicro is developing new server platforms incorporating the NVIDIA Grace CPU Superchip,” said Charles Liang, President and CEO at Supermicro. “Leveraging our innovative Building Block Solutions, we plan to offer a wide range of servers optimized for specific workloads. Based on our green computing DNA, Supermicro offers the most advanced energy-efficient servers available to meet demanding data center requirements. Our first-to-market new server systems can quickly address the rapidly expanding AI, HPC, and data analytics markets.”

HPC, AI

Supermicro will use the NVIDIA Grace CPU Superchip in systems that are designed for HPC, AI, and data analytics applications. The NVIDIA Grace CPU Superchip, in particular, interfaces with the latest PCIe Gen5 protocol, allowing for optimal 400Gb/s connectivity with the NVIDIA Quantum-2 InfiniBand platform or NVIDIA Spectrum Ethernet, including NVIDIA ConnectX-7 smart network interface cards and NVIDIA BlueField-3 DPUs, enabling secured and accelerated HPC and AI workloads.

For many years, Supermicro has worked with NVIDIA to provide the most powerful graphics, HPC, and AI accelerators to a wide spectrum of users. With the new NVIDIA Grace CPU Superchip, the entire NVIDIA software stack, including NVIDIA AI Enterprise, would run seamlessly on upcoming Supermicro servers. Supermicro will also continue to support NVIDIA Omniverse Enterprise, helping to accelerate 3D design collaboration and simulation of industrial virtual worlds and digital twins.