End-to-End AI Supercomputing Platform NVIDIA HGX H100/H200 8-GPU server This 7U dual-socket server is powered by 5th Gen Intel Xeon® Scalable Processors and eight NVIDIA H100/H200 Tensor Core GPUs. Designed to acceler

End-to-End AI Supercomputing Platform

NVIDIA HGX H100/H200 8-GPU server

This 7U dual-socket server is powered by 5th Gen Intel Xeon® Scalable Processors and eight NVIDIA H100/H200 Tensor Core GPUs. Designed to accelerate the development of AI and data science, HGX3 offers a dedicated one-GPU-to-one-NIC topology and supports up to eight NICs for the highest throughput during compute-intensive workloads. To reduce operating costs, the HGX3 | ESC N8-E11/ESC N8-E11V is engineered to provide effective cooling and innovative components that deliver thermal efficiency, scalability, and unprecedented performance.

KEY FEATURES

PCIe 5.0 Ready

Faster storage, graphics and networking capabilities

PCI Express® (PCIe®) 5.0 delivers 32 GT/s bandwidth, which is twice the speed of PCIe 4.0, and offers lower power consumption, better lane scalability and backwards compatibility. These servers are PCIe 5.0 ready, with scalable designs to satisfy the increasing workloads or modern data centers.

System Layout

Block Diagram

One-GPU-to-1-NIC Topology

Supports 8 NICs and 8 GPUs in one system

A dedicated 1-GPU-to-1-NIC topology that allows the ESC N8A-E12 to support up to 8 NICs, delivering the highest throughput for compute-intensive workloads.

Modular Design with Reduced Cable Usage

Easy troubleshooting and thermal optimization

The modular design greatly reduces the amount of cables used, resulting in a shortened system assembly time and avoiding cable routing, as well as improving thermal optimization by lowering the risk of choked airflow.

Advanced NVIDIA Technologies

The full power of NVIDIA GPUs, DPUs, NVLink, NVSwitch, and networking

The HGX3 accelerates the development of AI and data science by incorporating fourth generation NVLink and NVSwitch technology, as well as NVIDIA ConnectX-7 SmartNIC. This also empowers GPUDirect® RDMA, storage with NVIDIA Magnum IO™, and NVIDIA AI Enterprise – the software layer of the NVIDIA AI platform.

Sustainability

Keep your data center green

Data centers can consume huge amounts of electricity, raising environmental concerns. These servers are designed to optimize both total cost of ownership (TCO) and total cost to the environment (TCE), including thermal design, power efficiency and advanced cooling solutions — reducing the impact of our products on the planet.

Optimized Thermal Design

A two-level GPU and CPU sled for thermal efficiency

The HGX3 features a streamlined design with dedicated CPU and GPU airflow tunnels for efficient heat dissipation. The two-level GPU and CPU sled design enhances thermal efficiency, scalability, and overall performance by allowing heat to be expelled into the surrounding ambient air – resulting in greater energy efficiency and overall system power savings.

4+2 Power Supplies

High level of power efficiency

To reduce operating costs, the HGX3 provides effective cooling and innovative components. The 80 PLUS Titanium power supplies come in a redundant 4+2 or 3+3 arrangement to efficiently deliver an abundance of power.

Comprehensive liquid-cooling solutions

Better data-center power efficiency

As liquid cooling is maturing in the age of supercomputing, HGX3 is purpose-built for sustainability and ready to support direct-to-chip (D2C) cooling by request for maximum data-center energy efficiency.

Serviceability

Improved IT operations efficiency