Quanta/QCT is the leading partner with NVIDIA in introducing the MGX architecture - an open and future compatible accelerated computing reference architecture designed to allow rapid adoption of key platform technologies including CPUs, GPUs and DPUs. The modular architecture consists of configurable bays that can house different modules to achieve desired configurations. This allows for future hardware solutions with multiple power distribution methods, cooling solutions, including hot or cold aisle configurations.

Quanta GH200 is the first server to introduce the NVIDIA® Grace™ Hopper™ Superchip in conjunction with NVIDIA® MGX™ architecture. The Superchip combines 72 Arm Neoverse cores connected by NVLink® chip to chip high bandwidth interconnect with the Hopper™ H100 GPU to deliver a coherent memory pool that excels at accelerating AI and high performance computing applications. The modular infrastructure is designed to support multiple system configurations and reduce time to market while providing a compatible platform for future CPU, GPU and DPU solutions.

NVIDIA GH200 Grace Hopper Superchip

The breakthrough processor for large-scale AI and high-performance computing (HPC) applications.

The World’s Most Versatile Computing Platform

The NVIDIA Grace Hopper™ architecture brings together the groundbreaking performance of
the NVIDIA Hopper™ GPU with the versatility of the NVIDIA Grace™ CPU in a single superchip,
connected with the high-bandwidth, memory-coherent NVIDIA® NVLink® Chip-2-Chip (C2C)
interconnect.
NVIDIA NVLink-C2C is a memory-coherent, high-bandwidth, and low-latency interconnect for
superchips. The heart of the GH200 Grace Hopper Superchip, it delivers up to 900 gigabytes per
second (GB/s) of total bandwidth, which is 7X higher than PCIe Gen5 lanes commonly used in
accelerated systems. NVLink-C2C memory coherency increases developer productivity,
performance, and the amount of GPU-accessible memory. GH200 can be easily deployed in
standard servers to run a variety of inference, data analytics, and other compute- and memoryintensive workloads.

QUANTA_GH200.pdf