Supermicro NVIDIA MGX™ Systems

Supermicro's 1U NVIDIA MGX systems feature up to 2 NVIDIA GH200 Grace Hopper superchips with 2 NVIDIA H100 GPUs and 2 NVIDIA Grace CPUs, each with 480GB of LPDDR5X memory for the CPU and 96GB of HBM3 or 144GB of HBM3e memory for the GPU.the high-bandwidth, low-latency, memory-coherent NVIDIA C2C connects CPU, GPU and memory at 900 GB/s - seven times faster than PCIe 5.0. The modular architecture offers multiple PCIe 5.0 x16 FHFL slots to accommodate DPUs for cloud and data management as well as expandability for additional GPUs, networking and storage.

With the 1U-2 node design featuring 2 NVIDIA GH200 Grace Hopper superchips, Supermicro's proven direct-to-chip liquid cooling solutions can reduce operating costs by more than 40% while increasing compute density and simplifying rack-scale deployment for Large Language Model (LLM). Cluster and HPC applications.

The 2U Supermicro NVIDIA MGX platform supports both NVIDIA Grace and x86 CPUs with up to 4 full-size data center GPUs, such as NVIDIA H100 PCIe, H100 NVL or L40S.it also offers three additional PCIe 5.0 x16 slots for I/O connectivity and eight hot-swap EDSFF storage bays.

Supermicro offers NVIDIA networking to secure and accelerate AI workloads on the MGX platform.this includes a combination of NVIDIA BlueField-3 DPUs, which provide 2x 200 Gbps connectivity to accelerate user-to-cloud and data storage access, and ConnectX-7 adapters, which provide up to 400 Gbps InfiniBand or Ethernet connectivity between GPU servers.

Developers can quickly leverage these new systems and NVIDIA software offerings for any industry workload.these offerings include NVIDIA AI Enterprise, enterprise-class software that powers the NVIDIA AI platform and optimizes the development and deployment of production-ready generative AI, computer vision, speech AI and more. In addition, the NVIDIA HPC software development kit provides the essential tools needed to advance scientific computing.
Every aspect of Supermicro NVIDIA MGX systems is designed to increase efficiency, from intelligent thermal design to component selection.nVIDIA Grace superchip CPUs have 144 cores and deliver up to twice the performance per watt compared to today's industry-standard x86 CPUs. Specific Supermicro NVIDIA MGX systems can be configured with two nodes in 1U, for a total of 288 cores on two Grace CPU superchips to deliver breakthrough compute densities and energy efficiency in hyperscale and edge data centers.

Build new solutions for accelerated infrastructures that enable scientists and engineers to focus on solving the world's most important problems with larger data sets, more complex models, and new generative AI workloads. In the same 1U chassis, Supermicro's dual NVIDIA GH200 Grace Hopper superchip systems deliver the highest performance for any application on the CUDA platform with significant speedups for AI workloads with high memory requirements. In addition to hosting up to 2 onboard H100 GPUs in a 1U form factor, the modular bays enable full-size PCIe expansion for current and future accelerated compute components, high-speed scale-out and clustering.

Supermicro NVIDIA MGX™ 1U/2U systems with Grace™ CPU Superchip and x86 CPUs are fully optimized to support up to 4 GPUs over PCIe without compromising I/O networking or heat dissipation. The ultimate building block architecture allows you to optimize these systems for a variety of accelerated workloads and domains, including AI training and inference, HPC, data analytics, visualization/Omniverse™ and hyperscale cloud applications.