AIR-520 Next-Gen Edge AI Server

The rise of generative AI is transforming industries, and the Advantech AIR-520 edge AI server is designed to meet these evolving demands. Featuring Phison-patented aiDAPTIV+ technology and powered by the AMD EPYC 7003 series processor, the AIR-520 seamlessly integrates SQ ai100 AI SSDs, NVIDIA RTX GPU cards, Edge AI SDK, and NVIDIA AI Enterprise. This all-in-one, ready-to-deploy solution empowers enterprises to fine-tune large language models (LLMs) at the edge with up to four GPU cards, keeping sensitive data secure while minimizing infrastructure costs.

Advantech Edge AI SDK includes the GenAI training studio, which is preloaded with Llama-2 models (13B/33B/70B) for applications ranging from chatbots to data analysis, making custom LLM model training and inference straightforward. The AIR-520 offers four tailored configurations: AIR-520-L13B, L33B, L70B, and L70B-Plus, ensuring scalability to meet diverse AI application needs.

Features

• 4U edge AI server with AMD EPYC 7003 processor: delivers up to 64 cores and 768 GB DRAM, ensuring robust server-grade performance for AI workloads
• Powerful expansion and graphics support: equipped with four PCIe x16 expansion slots and a 1,200 W PSU for up to 700 W graphic card TDP, supporting high-performance NVIDIA RTX GPUs
• Advanced generative AI capabilities: supports fine-tuning and inference for LLMs with models like Llama-2 (13B, 33B, 70B) in a secure, edge-based environment
• Flexible model training options: GenAI training studio and Edge AI SDK offer a seamless experience for training and evaluating custom LLMs
• Customizable configurations are available in four options: AIR-520-L13B, L33B, L70B, and L70B-Plus, catering to a range of enterprise needs