AMD Launches Ryzen AI Halo Developer Platform With 192GB Memory

AMD launched the Ryzen AI Halo Developer Platform on Oct. 12, combining a 16-core processor with up to 192GB of unified memory for local artificial intelligence development.

The compact system uses AMD’s Ryzen AI Max+ PRO 495 with 16 Zen 5 CPU cores and 32 threads. Its shared LPDDR5x memory pool can reach 192GB, with up to 160GB assigned to the integrated GPU for graphics and AI workloads.

The platform is designed to support language models with 300 billion parameters or more at 4-bit quantization. The 4-bit format reduces the amount of memory required to run a model, while the unified-memory design lets the CPU, GPU and neural-processing unit draw from the same pool.

The system targets local AI development, including agent workflows, image and video generation, music tools and 3D applications. Graphics are handled by an integrated Radeon 8065S GPU with 40 compute units based on the RDNA 3.5 architecture.

The processor’s XDNA 2 neural-processing unit provides up to 55 trillion operations per second. Total AI performance is listed at up to 61 FP16 teraflops, while memory bandwidth reaches 273GB per second.

Other specifications include a 2TB M.2 SSD, 10Gbps Ethernet, Wi-Fi 7, PCIe 4.0 and a 120-watt thermal design power. Linux and Windows 11 are supported operating systems.

The platform also includes the Ryzen AI Developer Center, preloaded software and models, guided AI Playbooks and validated software configurations that can be updated or restored.

The Ryzen AI Halo platform is available to order through Micro Center. Windows-based systems using Windows Project Zenith are listed as coming soon.