Delivering Optimized AI Performance for Enterprise Applications


Powered by AMD EPYC™ Platforms

and AMD Instinct™ MI350P PCIe® Cards


Delivering Optimized AI Performance for Enterprise Applications


Powered by AMD EPYC™ Platforms and AMD Instinct™ MI350P PCIe® Cards

Upgrade Your AI Infrastructure with Advantech and AMD

As organizations adopt AI, many discover that upgrading on-prem infrastructure can require expensive redesigns in power and cooling. Advantech and AMD offer a smarter path. The AMD Instinct™ MI350P is a 10.5" FHFL dual-slot PCIe® card designed for standard air-cooled servers. It delivers the AI performance your business demands without the cost or complexity of purpose-built AI systems.

Core Advantages

Efficiency at Scale

Connects via x16 PCIe Gen 5 host interface to deliver up to 128 GB/s bandwidth.

Seamless Integration

Designed for the PCIe® form factor to drop into mainstream air-cooled servers without specialized cooling or rack redesigns.

Power Flexibility

600W total board power (TBP), configurable to 450W to meet specific power or cooling constraints.

Innovative Technology

Native Hardware

Support

Built-in acceleration for MXFP4, MXFP6, and MXFP8 data formats to maximize AI throughput. 



Advanced Security & Confidential AI

Features Device Secure Boot, Secure Update, and Secure Recovery to ensure firmware integrity.

Hardware-Level GPU Partitioning

Isolated Hardware Partitions: Support up to 4 isolated GPU physical partitions.



Why Advantech Edge Server for the AMD Instinct™ Platform

Assured Reliability Industrial DNA

Wide Operating Temperature

0 ~ 40 operation

Anti-Virbration

Design engineered for harsh

industrial environments

Industrial-Grade Design

Enhanced protection for
reliable operation


New Product Highlights

Outstanding Performance



  • Equipped with AMD EPYC™ 9005/9004 Series processors™
  • TDP up to 400W CPU

High-Density Expansion in a 4U MGX Platform



  • Supports up to 8 double-slot expansion cards in a 4U chassis (up to 600W TDP each)


ROCm Optimized


  • ROCm is optimized for Generative AI and HPC applications, and it is easy to migrate existing code into ROCm software


High-Density E1.S Storage Design


  • Supports up to 16 front-access E1.S SSDs

Accelerating AI from Edge to Cloud with SKY-642E3 and AMD Instinct™ MI350P PCIe® GPUs

AI-optimized infrastructure built for Generative AI, Agentic AI, and HPC workloads.

Key Applications

Achieve Success with Advantech!

We offer a team with expertise to service your needs! Contact a personal consultant now!

Contact Us

Follow Advantech