Edge AI server solution for generative AI

     Advantech has announced a breakthrough edge AI server solution that utilizes Phison's patented aiDAPTIV + technology. The AIR-520 Edge AI server is powered by AMD EPYC 7003 series processors and integrates SQ ai100 AI ssd, NVIDIA RTX GPU card, Edge AI SDK and NVIDIA AI Enterprise. Provides a ready to deploy solution.

     Generative AI tools such as the LLM have transformed enterprise knowledge management by automating data organization, retrieval and analysis, increasing productivity and improving decision-making. Custom llm improves accuracy, while edge training increases data privacy, although it may cost more. The solution supports LLM fine-tuning of 1-4 GPU cards and the Square ai100 AI ssd, enabling enterprises to cost-effectively train LLMS while keeping sensitive data at the edge secure.

     The company offers four options: the Air-520 l13b/L33B L70B and L70B - in addition, tailored for different sizes and applications. The l13b is ideal for real-time applications such as chatbots and language translation. The l33b is suitable for more complex tasks, increasing the productivity and innovation of content creation. The l70b excels in complex data analysis and decision-making in specialized areas. In addition, the L70B-Plus is equipped with the NVIDIA AI Enterprise software platform, which provides an end-to-end, reliable and optimized AI sdk, and provides long-term support and expert consulting services for efficient business application deployment.

     All solutions include SQ ai100 AI solid-state drives using Phison's aiDAPTIV+ technology. These SSDS act as extensions to the GPU vRAM, allowing the system to fine-tune llm with minimal GPU cards. This approach not only mitigates budget barriers, but also makes Edge AI servers more compact than traditional mainframe servers. The air-520 Edge AI server is designed for a variety of edge AI applications. It is about the size of a desktop computer and can be mounted on a rack with the appropriate accessories. Its low profile allows for easy deployment of edge AI fine-tuning environments, eliminating space and maintenance concerns.

     In addition to LLM fine-tuning capabilities, the company also offers an edge AI SDK with GenAI training studios preloaded with Llama-2 13 b / 33 b / 70 b models for applications such as chatbots and data analytics. This simplifies and speeds up customer-specific llm model training and inference evaluation on the air-520. In addition, its DeviceOn provides OTA software/container updates and remote management, enabling efficient edge AI orchestration and long-term maintenance.