QCT and Supermicro are the first to use the server specification to support more than 100 system configurations to accelerate AI, HPC, Omniverse workload
COMPUTEX - To meet the diverse accelerated computing needs of data centers around the world, NVIDIA today released the NVIDIA MGX™ server specification, which is System builders provide a modular reference architecture that can quickly and cost-effectively build more than 100 server variants to suit a wide range of artificial intelligence, high-performance computing, and omniverse applications.
ASRock Rack, ASUS, GIGABYTE, Pegatron, QCT and Supermicro will use MGX to cut development costs by up to three-quarters and development time by two-thirds to just six months.
“Enterprises are looking for more accelerated computing options as they build data centers that meet their specific business and application needs,” said Kaustubh Sanghani, vice president of GPU products at NVIDIA. time and money. "
With MGX, manufacturers start with a base system architecture optimized for accelerated computing in server chassis, then select their GPUs, DPUs and CPUs. Design variants can address unique workloads such as HPC, data science, large language models, edge computing, graphics and video, enterprise AI, and design and simulation. Multiple tasks such as AI training and 5G can be handled on a single machine, while upgrading to future generations of hardware can be done smoothly. MGX can also be easily integrated into cloud and enterprise data centers.
Partnering with industry leaders
QCT and Supermicro will be first to market with the MGX design available in August. Supermicro's ARS-221GL-NR system, announced today, will include the NVIDIA Grace™ CPU superchip, while QCT's S74G-2U system, also announced today, will use the NVIDIA GH200 Grace Hopper superchip.
Additionally, SoftBank Corp. plans to roll out multiple hyperscale data centers across Japan and use MGX to dynamically allocate GPU resources between generating AI and 5G applications.
"Building the right infrastructure at the right cost is one of the biggest challenges network operators face as generative AI permeates the way businesses and consumers live," said Junichi Miyakawa, President and CEO of SoftBank Corp. Hopefully NVIDIA MGX can meet these challenges and allow for multipurpose AI, 5G, and more, depending on real-time workload requirements.”
Different Designs for Different Needs
Data centers are increasingly required to meet increasing computing power and reduce carbon emissions in response to climate change, all while reducing costs.
NVIDIA's accelerated computing servers have a long history of delivering exceptional computing performance and power efficiency. The MGX's modular design now enables system builders to more efficiently meet each customer's unique budget, power delivery, thermal design and mechanical requirements.
Multiple form factors for maximum flexibility
MGX is available in different form factors and is compatible with current and future generations of NVIDIA hardware, including:
- Chassis: 1U, 2U, 4U (air-cooled or liquid-cooled)
- GPU: Complete NVIDIA GPU portfolio including the latest H100, L40, L4
- CPU：NVIDIA Grace CPU Superchip、GH200 Grace Hopper Superchip、x86 CPU
- Network: NVIDIA BlueField ® -3 DPU, ConnectX ® -7 Network Adapter
MGX differs from NVIDIA HGX ™ by offering flexible, multi-generational compatibility with NVIDIA products to ensure system builders can reuse existing designs and easily adopt next-generation products without costly redesigns. In contrast, HGX is an NVLink®-connected multi-GPU backplane that is custom-built for scaling to create the ultimate AI and HPC system.
Software that Pushes Acceleration Further
In addition to the hardware, MGX is supported by NVIDIA's complete software stack, enabling developers and enterprises to build and accelerate AI, HPC and other applications. This includes NVIDIA AI Enterprise, the software layer of the NVIDIA AI platform with more than 100 frameworks, pretrained models and development tools that accelerate AI and data science to fully support enterprise AI development and deployment.
Compatible with Open Compute Project and Electronic Industries Alliance server racks, the MGX can be quickly integrated into enterprise and cloud data centers.
Watch NVIDIA founder and CEO Jensen Huang discuss the MGX server specification during his keynote at COMPUTEX.
