AMD has stepped up its push into the rapidly growing local AI PC market with its new Ryzen AI Max+ PRO 495, code-named Gorgon Halo. The chip is attracting attention for its massive 192GB of LPDDR5X unified memory, allowing compatible systems to handle significantly larger AI models locally without depending entirely on cloud computing. AMD says its Ryzen AI Halo platform can run models with up to 300 billion parameters.
The Ryzen AI Max+ PRO 495 is built on AMD’s Zen 5 architecture and features 16 CPU cores and 32 threads, with boost clocks reaching up to 5.2GHz. It also integrates Radeon 8065S graphics with 40 compute units based on RDNA 3.5, along with an XDNA 2 NPU capable of up to 55 TOPS of AI performance. The platform can deliver up to 61 FP16 TFLOPS of GPU compute performance.
The biggest feature, however, is the unified memory architecture. The 192GB configuration can allocate a large portion of memory to GPU workloads, making the system particularly useful for generative AI, LLM inference, image generation, software development and professional content creation. AMD is positioning this capability against emerging AI platforms such as NVIDIA’s RTX Spark.
AMD also released new ComfyUI generative-AI benchmark results on October 5. In AMD’s testing, the Ryzen AI Max+ PRO 495 was compared with Intel’s Core Ultra X9 388H system. Performance advantages varied widely by workload, ranging from 1.1x to 32.2x, although the highest figure was identified as an outlier by Tom’s Hardware.
The first Gorgon Halo-powered systems are already reaching the market, with some high-end configurations costing around $7,000 or more. As NVIDIA prepares its RTX Spark push, the competition for powerful local-AI hardware is getting increasingly intense.
Comments