Lanner Electronics has announced its next-generation edge AI server, the ECA-6051, powered by the NVIDIA MGX reference architecture. Designed to accelerate AI training and inference in 5G infrastructure, the ECA- 6051 is a 2U short-depth edge AI server featuring the NVIDIA GH200 Grace Hopper Superchip and support for multiple GPUs. The ECA-6051 aims to enable low-latency multi-access edge computing applications, including video transcoding, factory visual inspection, and RAN intelligent control, in private 5G networks.

To meet the increasing need for AI acceleration in 5G infrastructure, theECA-6051 is equipped with CPU boards using an NVIDIA GraceCPU with 72 Arm Neoverse V2 cores, plus a maximum of 1,536GB DDR5 memory. Featuring a front-access, 420mm-depth short-chassis and redundant power supplies, the ECA-60 51 is specifically designed for deployment in space-constrained 5G edge network locations. It accelerates the performance and minimizes the deployment time of AI-driven applications while minimizing latency caused by data transmission to centralized servers in the cloud.

Configured with up to 3 PCIe 5.0 x16 slots, the modular ECA-6051 can accommodate 3 PCIe cards, including the NVIDIA L40S GPU, NVIDIA H100Tensor CoreGPU, NVIDIA Bluefield-3 DPU, and NVIDIA ConnectX network adapters. With support for multiple GPUs and DPUs, this edge AI server dramatically enhances AI and networking performance at the 5G edge, providing a scalable, high-performance platform for telecommunications operators to construct and deploy 5G radio access networks that are open, efficient, and secure.