AMD has announced the Radeon AI PRO R9700 GPU, featuring RDNA 4 architecture and 32GB of memory

Unveiled during Computex 2025, the AMD Radeon AI PRO R9700 GPU features RDNA 4 technology optimized for AI-driven workstations. With 32GB of GDDR6 memory, it boasts 96 TFLOPS of Peak Half-Precision and an impressive 1531 TOPS of INT4 Sparse AI performance, resulting in performance improvements of up to 2X compared to the prior generation's AMD Radeon PRO W7800 32GB GPU.

The 32GB VRAM of the Radeon AI PRO R9700 makes it particularly well-suited for advanced local text-to-image AI models and large language models such as DeepSeek R1 Distill Qwen 32B Q6 and Mistral Small 3.1 24B Instruct 2503 Q8. With enhanced AMD ROCm support on the Radeon, it will accommodate a wider variety of AI and compute tasks.

The initial chart in the presentation for the new AI workstation GPU illustrates a comparison of average tokens per second performance against the GeForce RTX 5080, exhibiting an increase of up to 496%. See the chart below for details.

The Radeon AI PRO R9700 is also designed to be scalable. A single workstation can support up to four GPUs, providing sufficient memory to manage models with 123 billion and 70 billion parameters, indicating significant AI capabilities. AMD anticipates launching the Radeon AI PRO R9700 in July 2025 (pricing details are yet to be announced), with various models from partners such as ASRock, ASUS, GIGABYTE, PowerColor, Sapphire, XFX, and Yeston set to be available.

The AMD Radeon AI PRO R9700 is a formidable tool designed for workflows centered around AI and productivity, delivering top-of-the-line performance for complex tasks. However, if your focus is on gaming performance—with smooth frame rates, breathtaking visuals, and next-generation features—the Radeon RX 9070 stands as the premier option. As always, for the latest news on hardware launches and industry developments, be sure to follow our dedicated hardware coverage.