Description
The new NVIDIA MGX Platform for versatile solutions
NVIDIA MGX GH200 is a part of the new NVIDIA MGX platform, providing versatile and highly flexible solutions. With MGX, OEM and ODM partners can build tailored solutions for different use cases while saving development resources and reducing time to market. The modular reference architecture allows for different configurations of GPUs, CPUs, and DPUs—including NVIDIA Grace, x86, or other Arm® CPU servers, and NVIDIA OVX™ systems—to accelerate diverse enterprise data center workloads.
Why NVIDIA MGX?
Great configurability
With unified modular layout, the NVIDIA MGX system can be configured in hundreds different ways under the same platform.
Cross-platform support
NVIDIA MGX systems can host any particular hardware - X86 AMD/Intel platforms, ARM CPUs and all current generation NVIDIA GPUs.
Futureproof solution
The MGX system's modular layout ensures compatibility for any future hardware generation and accelerating market-to-deployment time.
Supermicro ARS-111GL-NHR MGX GH200 system
ARS-111GL-NHR is a NVIDIA MGX GH200 Grace Hopper Superchip system supporting NVIDIA BlueField-3 or NVIDIA ConnectX-7, suitable for High Performance Computing (HPC), AI/Deep Learning Training and Inference, Large Language Model (LLM) and Generative AI applications.
NVIDIA H200 GPU
NVIDIA H200 GPU with 141GB HBM3 Memory and 4.8TB/s bandwidth
GH200 superchip
NVIDIA Grace Hopper™ Superchip (Grace CPU and H200 GPU)
Large system memor
Up to 576GB of coherent memory per node for LLM applications
High-density system
High density 1U GPU system with Integrated NVIDIA® H200 GPU
C2C NVLink
High-bandwidth, low-latency interconnect between CPU and GPU
High-speed interface
2x PCIe 5.0 x16 slots supporting NVIDIA BlueField®-3 or ConnectX®-7
Powered by NVIDIA GH200
NVIDIA MGX GH200 System is powered by the Hopper architecture, used in the H200 GPU.
ARS-111GL-NHR Specificaion
Format | High-density 1U GPU server |
CPU | 72-core Grace Arm Neoverse V2 CPU + H100 Tensor Core GPU in a single chip |
GPU | NVIDIA H100 Tensor Core GPU with 96GB of HBM3 or 144GB of HBM3e |
Memory | 480GB integrated LPDDR5X with ECC |
Drives | 8x hot-swap E1.S drives and 2x M.2 NVMe drives |
Networking | 3x PCle 5.0 x16 slots supporting NVIDIA BlueField-3 or ConnectX-7 |
Interconnect | NVLink-C2C with 900GB/s for CPU-GPU interconnect |
Cooling | Air-cooling, 8 hot-swap high-flow fans |
Power | 2x 2000W Redundant Titanium Level power supplies |
Reviews
There are no reviews yet.