MSI is another of the multiple manufacturers that have been present at the GTC 2025, the conference that Nvidia has held this week in the Californian city of Sans José and that has left us great novelties in graphic technologies, chips and accelerators for ia.
The Taiwanese firm has presented new High performance servers, AI 4U and 2U platforms Based on the reference architecture NVIDIA MGX presented at the conference. This platform aims to bring accelerated computing to any data center with modular servers designs that can be used for a wide variety of use cases, from remote visualization to perimeter supercomputing.
It is the approach chosen by MSI for its new AI servers based on the MGX and designed to meet the various demands of AI, HPC and applications with intensive data use. They offer scalable performance to meet the needs of business and cloud data centers, taking advantage of modular functions to optimize the strengths of AI workloads.
“The adoption of AI has become a fundamental objective for business data centers, since organizations seek to take advantage of advanced artificial intelligence capabilities to obtain competitive advantages”explained Danny Hsu, general director of the MSI Business Platform Solutions Group.
MSI servers with nvidia mgx
Modular architecture From NVIDIA MGX allows different configurations of GPU, CPU and DPU and taking advantage of this type of flexible format, the new MSI AI platforms feel the foundations for a scalable and vision of the future infrastructure. The CG480-S5063 servers of 4U and CG290-S3063 of 2U offer a robust and scalable base to meet the changing demands of GPU technologies and AI servers, promoting innovation and long-term adaptability for companies and data centers.
The AI server platform CG480-S5063 DE 4U It is an avant -garde solution designed for advanced AI work loads, including large -sized language models (LLM), deep learning training and complex data analysis. Promoted by NVIDIA MGX architecture, it has two Intel Xeon 6 and eight Double Width FHFL GPUR processors, compatible with NVIDIA H200 NVL and NVIDIA RTX PRO 6000 BLACKWELL SERVER EDITION with power capabilities up to 600 W. It has 32 Dimm DDR5 slots and twenty bays PCIE 5.0 E1.S NVME.
The second of the AI server platforms is the CG290-S3063 2Udesigned to meet the diverse needs of IA and HPC workloads. Also based on NVIDIA MGX architecture, it incorporates an Intel Xeon 6 socket processor and four double -wide FHFL GPU slots, which offer compatibility with 600 W rear I/O compatibility and 400 W reate in two sku options. Equipped with PCIE 5.0 expansion slots, eight bays for 2.5 -inch units and two M.2 NVME slots, this server offers flexible configurations and great processing power for applications that range from small -scale inference to large -scale AI training.
