Server Technology Highlights – HGX, EPYC, and Xeon at a Glance

In this category, we showcase the technical highlights of our server lineup: platforms that are defined not by a specific use case or height unit, but by a unique technical foundation. Currently, the focus here is on our NVIDIA HGX platform for large-scale AI training; we are continuing to expand our offerings to include additional form factors and technologies.

NVIDIA HGX Servers: SXM Platform for Large-Scale AI

NVIDIA HGX servers rely on the SXM architecture instead of traditional, card-based PCIe GPUs. Up to eight NVIDIA H200 GPUs are mounted on a single baseboard and communicate directly via NVLink, at up to 900 GB/s per GPU. They are connected via the PCIe lanes of two AMD EPYC or Intel Xeon Scalable processors. ECC memory ensures stability during continuous operation. If your workload requires fewer than eight graphics cards, you’ll find the right alternative in our GPU servers.

Your Path to the Right Server Configuration

In the Server Configurator, you can customize your system to meet your specific needs. If you’re still unsure which platform is right for your project, our server buying guide will help you with concrete recommendations. Before shipment, every system undergoes several days of quality control and stress testing under real data center conditions.

Business customers also benefit from our B2B support, which includes 24-hour on-site service provided by certified technicians.