Up to 192GB Unified Memory·Supports Models with Over 300 Billion Parameters

AMD has introduced the next-generation 'Ryzen AI Max PRO 400 Series' processor for building enterprise local AI environments, moving forward with its strategy to capture the agentic AI market.

AMD recently unveiled the Ryzen AI Max and Ryzen AI Max PRO 400 Series processors for enterprise PCs and workstations, and announced a platform strategy supporting enterprises in developing and operating local AI agents.

AI agents are a form of AI capable of performing multi-stage tasks by linking multiple applications and data, unlike generative AI focused primarily on question-and-answer exchanges. Enterprises are expanding the adoption of AI agents for business automation and productivity improvement; however, cost, data security, and cloud dependency have been identified as challenges.

Supporting Large-Scale AI Models with Up to 192GB Memory

The Ryzen AI Max PRO 400 Series features up to 16 'Zen 5' CPU cores, an NPU with 55TOPS performance, and up to 40 GPU compute units.

Notably, it supports up to 192GB of unified memory, with up to 160GB available for GPU workload allocation. AMD explains that this enables execution of large-scale AI models with over 300 billion parameters based on 4-bit quantization in a local environment without cloud migration.

Enterprise AI agents require high memory capacity as they utilize not only large language models (LLMs) but also in-house documents and databases, development code, and retrieval-augmented generation (RAG) systems.

AMD states that the new processor is designed to process these complex workloads in a local environment.

Emphasizing Hybrid Strategy Combining Cloud and Local AI

AMD presented a hybrid strategy employing both local and cloud approaches as an AI infrastructure operation method.

For complex reasoning or cases requiring ultra-large models, cloud AI is leveraged, while tasks handling sensitive information such as in-house documents or development materials are processed in a local AI environment.

The company explains that utilizing local AI can reduce external data transmission and cut continuous cloud inference costs.

Additionally, it supports major AI development frameworks including PyTorch, vLLM, llama.cpp, Ollama, ComfyUI, and LM Studio, providing an integrated development environment capable of performing AI model development, testing, and deployment.

AMD explains that through the Ryzen AI Max PRO 400 Series, it plans to provide enterprises with more efficient options for developing and operating AI agents.