AMD Announces Radeon AI PRO R9700 – AI-Centric GPU for Professional Workstations

Live from Taipei, Taiwan – Computex 2025

While all eyes are on next-gen gaming this Computex, AMD is also making serious moves in the professional and AI workstation space. The newly announced Radeon AI PRO R9700 is built for local inference, AI model fine-tuning, and complex creative workloads, all on-device.

Engineered with 128 next-gen AI accelerators, the R9700 is powered by AMD’s RDNA 4 architecture, offering up to 4X the AI throughput of its predecessor. With 32GB of GDDR6 VRAM, PCIe Gen 5.0 support, and a focus on multi-GPU scalability, it’s designed to meet the evolving needs of professionals working with large language models (LLMs), real-time rendering, and AI-enhanced creative tasks.

Key Specs – Radeon AI PRO R9700:

• 128 RDNA 4 AI Accelerators

• 32GB GDDR6 Memory

• PCIe 5.0 Platform

• Up to 96 Half-Precision TFLOPS

• Up to 1531 TOPS (INT4 Sparse)

• 300W TDP

• Full ROCm Support on Linux (Windows support coming soon)

  
  

AMD is positioning the R9700 for edge AI deployments and high-performance workstations that demand both on-prem control and scalable compute capacity. Multi-GPU configurations are a key focus here, enabling massive VRAM pools and compute power for parallel workloads, simulations, and generative AI pipelines.

Compared to the previous-gen Radeon PRO W7800, AMD claims up to 2x better performance in key AI inference scenarios such as DeepSeek R1 and Distill Llama 8B.

The AMD Radeon AI PRO R9700 will be available starting July 2025, with full spec sheets and benchmarks expected closer to launch.

Stay tuned as we cover the rest of AMD’s announcements live from Computex, including hands-on coverage and deeper insights into AI-centric hardware trends.