Supermicro drives advanced AI capabilities to edge computing

Using Supermicro application-optimised servers with NVIDIA GPUs makes it easier to fine-tune pre-trained models and for AI inference solutions to be deployed at the edge where the data is generated, improving response times and decision-making.

“Supermicro has the broadest portfolio of Edge AI solutions, capable of supporting pre-trained models for our customers’ edge environments,” said Charles Liang, President and CEO of Supermicro. “The Supermicro Hyper-E server, based on the dual 5th Gen Intel® Xeon® processors, can support up to three NVIDIA H100 Tensor Core GPUs, delivering unparalleled performance for Edge AI. With up to 8TB of memory in these servers, we are bringing data centre AI processing power to edge locations. Supermicro continues to provide the industry with optimised solutions as enterprises build a competitive advantage by processing AI data at their edge locations.”

With these server advancements, users no longer need to send data back to the cloud for processing, only to retrieve the information back to the edge, where it’s required. Customers can now use pre-trained large language models (LLMs), optimised for performance and available with NVIDIA AI Enterprise at their edge locations where the data is needed for accurate, real-time decision-making close to the data origination.

“Businesses across industries, including healthcare, retail, manufacturing, and auto, are increasingly looking to leverage AI at the edge,” said Kevin Connors, vice president of partner alliances at NVIDIA. “The new Supermicro NVIDIA-Certified Systems, powered by the NVIDIA AI platform, are built to deliver the highest-performing accelerated computing infrastructure, as well as NVIDIA AI Enterprise software to help run edge AI workloads.”

For example, Supermicro’s Hyper-E server, the SYS-221HE, is optimised for edge training and inferencing and supports dual socket CPUs in a short-depth, front I/O system. The system holds up to 3 double-width NVIDIA Tensor Core GPUs, including the NVIDIA H100, A10, L40S, A40, and A2 GPUs. These GPUs give the Supermicro Hyper-E sufficient computing power to process AI workloads at edge environments where data is collected, analysed, and stored. The Supermicro SYS-221HE is available with front or rear servicing options, allowing this server to be installed in various environments. As an example of the power and flexibility of the Supermicro Hyper-E server, partners such as Eviden are creating Edge AI solutions that enhance the customer experience while shopping in traditional retail outlets.

“Eviden’s AI-powered retail solution, based on the Supermicro edge systems and NVIDIA technologies, completely transforms the way people navigate through and interact with spaces such as stores. This enhances the shopping experience by offering customers interactive and personalised shopping through 3D models of the stores and interactive chatbots that can communicate appropriate information. The blend of lifelike facial animations, advanced speech recognition, and 3D modelling can make virtual shopping nearly as tangible as visiting a physical store”, said Jacque Istok, CEO of StoreGenius, specialising in smart applications for the retail industry.