Back

NVIDIA DGX, MGX, EGX and HGZ platforms

NVIDIA’s product lines, including DGX, MGX, EGX, and HGX, cater to various aspects of computing, from AI and deep learning to modular server design and edge computing. Each of these platforms is designed with specific use cases in mind, leveraging NVIDIA’s expertise in GPU technology to address the unique needs of different computing environments. While DGX is focused on deep learning and AI applications, MGX offers a modular and flexible architecture for a wide range of computing needs, including future-proofing for data centers. EGX targets edge computing with a combination of powerful compute and remote management capabilities, and HGX provides a high-performance computing platform for AI and HPC applications. Each platform serves distinct market segments, reflecting NVIDIA’s broad approach to addressing the computational demands across various industries.

NVIDIA DGX

The NVIDIA DGX platform is specifically built for enterprise AI, combining NVIDIA’s software, infrastructure, and expertise. It is designed to accelerate deep learning applications using General-Purpose computing on Graphics Processing Units (GPGPU). DGX systems typically feature a high-performance x86 CPU architecture, with models like the DGX A100 and DGX Station A100 utilizing AMD EPYC CPUs. These systems are equipped with 4 to 16 NVIDIA Tesla GPU modules, large heatsinks, and powerful fans to manage thermal output efficiently. DGX systems are ideal for tackling massive datasets and are used in environments where deep learning applications are critical[1][9].

NVIDIA MGX

The NVIDIA MGX platform offers a modular architecture for accelerated computing, aiming to meet the diverse needs of data centers worldwide. It provides system manufacturers with a modular reference architecture, enabling quick and cost-effective construction of high-performance computing and Omniverse applications. MGX differs from HGX by offering flexible, multi-generational compatibility with NVIDIA products, allowing system builders to reuse existing designs and easily upgrade to future hardware generations. MGX supports various form factors and is compatible with a wide range of NVIDIA hardware, including CPUs and GPUs. It is designed for a variety of workloads, including HPC, data science, large language models, edge computing, and enterprise AI[2][6][10][13].

NVIDIA EGX

The NVIDIA EGX platform is designed for accelerated computing at the edge, combining powerful compute capabilities with remote management technologies. It is a highly flexible server reference that can be configured to accelerate multiple professional visualization workloads. EGX servers combine high-end NVIDIA GPUs with virtual GPU (vGPU) software and high-performance networking, making them suitable for AI at the edge, IoT, and real-time processing applications[3][7][11].

NVIDIA HGX

The NVIDIA HGX platform is an AI supercomputing platform designed for high-performance computing (HPC) and AI applications. It features the HGX H100, a key GPU server building block powered by the NVIDIA Hopper Architecture. The HGX H100 platform includes eight H100 Tensor Core GPUs connected by third-generation NVSwitches, offering a fully non-blocking switch that connects all GPUs concurrently. This platform is optimized for dense HPC deployment and is designed to deliver high performance with low latency, securely integrating a full stack of capabilities from networking to compute at data center scale. HGX is ideal for scientific research, AI research, and environments requiring the ultimate in AI and HPC systems[4][8].

Citations:

[1] https://www.nvidia.com/en-us/data-center/dgx-platform/

[2] https://www.nvidia.com/en-us/data-center/products/mgx/

[3] https://www.nvidia.com/en-us/data-center/products/egx/

[4] https://www.nvidia.com/en-us/data-center/hgx/

[5] https://www.nvidia.com/en-gb/data-center/dgx-systems/

[6] https://nvidianews.nvidia.com/news/nvidia-mgx-server-specification

[7] https://www.nvidia.com/en-au/design-visualization/egx-server/

[8] https://developer.nvidia.com/blog/introducing-nvidia-hgx-h100-an-accelerated-server-platform-for-ai-and-high-performance-computing/

[9] https://en.wikipedia.org/wiki/Nvidia_DGX

[10] https://www.supermicro.com/en/accelerators/nvidia/mgx

[11] https://www.nvidia.com/en-us/design-visualization/egx-graphics/