;ESC-N4A-E11 | 4x A100 GPU AI Server with AMD EPYC | ASUS - RackmountNTS

ASUS 3U ESC N4A-E11

Data center | Enterprise | College
  • GPU: 4× NVIDIA® A100 Tensor Core GPUs (HGX A100 platform)
  • CPU: AMD EPYC™ 7003 Series Processor
  • Storage: 4× storage bays, 1× M.2 slot
  • Expansion: 3× PCIe® 4.0 slots
  • Power Supply: 80 PLUS® Titanium certified
  • Management: ASUS ASMB10-iKVM
  • Networking: OCP 3.0 support

Configure & Buy Features

AI Server for Simulation and Data Analytics

 

 

ASUS ESC N4A-E11 is a AMD EPYC 7003 server for AI supercomputing performance and leverages the NVIDIA HGX A100 baseboard designed to efficiently accelerate large complex AI workloads and HPC performance, with four Multi-Instance GPU (MIG)-capable Tensor Core GPUs to fully interconnect with up to 320 GB of total GPU memory. The system’s optimized thermal design enables support for an up to 400 W GPU* to deliver unprecedent performance for data analytics, AI and simulation.

 

 

 

 Key Platform Highlights

 

  • NVIDIA® HGX A100 with four Tensor Core GPUs delivers 80 teraFLOPS of FP64 for HPC workloads
  • Fueled by a single AMD EPYC 7003 series processor supporting TDP of 280 W
  • Optimized thermal design enables 400 W GPUs for better AI, simulation and data-analytics performance
  • Direct GPU-to-GPU interconnect via NVLink delivers up to 200 GB/s bandwidth for efficient scaling
  • Energy-efficient design with independent CPU and GPU-airflow tunnels for power savings
  • Three PCIe® 4.0 expansion slots support 200 GB Mellanox HDR InfiniBand smart network interface controller
  • High-throughput performance via NVIDIA GPU Direct Storage with four 3.5-inch drive bays and one onboard M.2
  • 3000 W 80 Plus Titanium power supply reduces operating costs and enables easier servicing
  • Onboard ASUS ASMB10-iKVM with ASPEED AST2600 controller for out-of-band management
  • Integrated PFR FPGA as the platform Root-of-Trust solution for firmware resiliency

 

 

The system overview features

 

Fully interconnected and Multi Instance GPU (MIG) capable with up to 320 GB of total GPU memory

NVIDIA NVLink

 

The A100 to A100 peer bandwidth is up to 200 GB/s bi directional, delivering more than 2X faster than PCIe Gen 4.0 x16 bus

AMD EPYC 7003 Series Processor

 

A single socket server supports TDP of 280W and 16 DIMM slots

Better Cooling System

 

Independent CPU and GPU airflow tunnel for sufficient heat dissipation

High speed Connectivity

 

Three PCIe 4.0 expansion slots supports 200GB Mellanox HDR InfiniBand NIC

 

 

NVIDIA GPUDirect Storage

 

Four 3.5 inch hot swap drive bays on front panel and one M.2 onboard

 

 

                     G4L3‑ZD1‑LAX5 GPU Server

Direct GPU-to-GPU Interconnect

 

The four A100 GPUs on the HGX baseboard of ESC N4A-E11 are directly connected through third-generation NVLink, enabling all GPUs to work as one on large AI models, with high-bandwidth communication for efficient scaling. The A100-to-A100 peer bandwidth is up to 200 GB/s bi-directional, for speeds that are up to 2X faster than a PCIe 4.0 x16 bus.

 

AMD EPYC 7003 Processors

 

Built for AI supercomputing, ESC N4A-E11 features a single-socket AMD EPYC 7003-series processor offering the highest performance for AI and HPC workloads and supports a TDP of 280 W and extensive memory capacity of 16 DIMM slots populated with 8-channel DDR4-3200 memory per processor.

 

 

 

NVIDIA GPUDirect Storage

 

Featuring four 3.5-inch hot swap drive bays supporting up to two NVMe SSDs on front panel and one M.2 onboard that enables extensive storage and high-throughput performance. ESC N4A-E11 supports GPUDirect Storage to provide a direct path between storage and GPU memory, avoiding extra copies through a bounce buffer in the CPU’s memory.

Titanium Power Supply

 

ESC N4A-E11 supports 3000 W 80 Plus Titanium redundant power supplies which are 96%+ efficient, directly reducing operating costs and offering easier servicing. The 1+1 power-supply design allows the server to keep working even if one power supply requires maintenance, enabling uninterrupted operation.

Enhanced Security

 

ASUS ESC N4A-E11 servers integrate PFR FPGA as the platform Root-of-Trust solution for firmware resiliency to prevent from hackers from gaining access to infrastructure. ASUS security solutions are fully compliant with the 2018 National Institute of Standards and Technology (NIST) SP 800 193 specification.

In addition, all ESC N4A-E11 servers also include support Trusted Platform Module 2.0 (TPM 2.0) to secure hardware through integrated cryptographic keys and offer regular firmware update for vulnerabilities.

ASUS ASMB10-iKVM

 

ASUS ASMB10-iKVM is the latest server-management solution from ASUS, built upon the ASPEED 2600 chipset running on the latest AMI MegaRAC SP-X. The module provides various interfaces to enable out-of-band server management through WebGUI, Intelligent Platform Management Interface (IPMI) and Redfish® API.

OPERATING SYSTEM
NTS AI Stack See what's inside each NTS AI package
Call for pricing