Advantech is demonstrating its GPU servers for AI and IoT applications at NVIDIA GTC 2025. With the complexity of AI, HPC and data analytics increasing exponentially over the next decade, scientists and developers will need advanced computing platforms with performance that is several orders of magnitude better than today to overcome these extraordinary challenges.
With the NVIDIA MGX reference architecture, system manufacturers can quickly and cost-effectively build more than 100 server variations. Advantech is releasing a series of NVIDIA MGX GPU servers designed to support a wide range of AI, high-performance computing, and NVIDIA Omniverse-based physical AI simulation applications. The Advantech SKY-622G4 is a 2U GPU server with dual Intel Xeon 6 processors, supporting up to four double-width NVIDIA GPUs for intensive AI and HPC workloads. It features 7 x PCIe 5.0 expansion slots, enabling seamless Smart NIC and DPU integration for improved networking and computing performance. Furthermore, the server is equipped with 8 x E1 storage bays and up to 6000W CRPS AC power supplies to meet the demands of AI-driven computing environments.
To fuel Omniverse applications, which need high-performance CPUs, multiple GPUs, and high-speed networking, the server is an OVS Certified ready server that supports up to four double-width GPU cards, including the NVIDIA RTX PRO 6000 Blackwell Server Edition, a powerful data centre GPU for AI and visual computing. It accelerates demanding enterprise workloads, including AI, scientific computing, graphics, and video applications. It supports MIG, allowing up to four fully isolated instances to power AI and graphics workloads. With 96GB of GDDR7 memory, it delivers twice the memory capacity and up to five times the performance of the L40S, making it an excellent choice for large-scale LLM inference applications.
The Advantech SKY-6 series GPU servers aim to transform enterprises by empowering them with intelligent connectivity, extending from IoT devices to cloud-based services. The SKY-6 series can scale out from a single GPU node to multiple GPU nodes when needed in industrial applications, particularly in areas such as healthcare and video streaming. With NVIDIA EGX platform support, the SKY-6 GPU servers provide a wide range of optimised and efficient solutions for intelligent edge applications. They provide seamless integration from the network edge to the public or private cloud and support evolving applications that enable an intelligent planet.
