We all know that the term “AI” is officially tiring, and that’s not a good thing, especially when you see hardware that is pretty capable of AI features, the ones that will shape a lot of things for tomorrow. With that, here at Computex 2025, AMD didn’t release one GPU but two—one for consumers and one for AI specifically, and that’s the AMD Radeon AI PRO R9700 Graphics.
Computex 2025 AMD Radeon AI PRO R9700
Make no mistake, this GPU is specifically designed for those seeking deep (see what we did there?) within the AI realms. It’s based on the AMD RDNA 4 AI Architecture with 32GB of GDDR6, 128 AI accelerators, supporting up to 96 TFLOPs peak half-precision and up to 1531 TOPs on INT4 Sparse. All that with a power consumption (TDP) of 300W. If you do want to compare to the AI-less AMD Radeon PRO W7800, the performance with the new one happens to be 2x faster than the older GPU, specifically tested in DeepSeek R1 Distill Llama 8B.
Again, it isn’t a GPU for the consumers, and while there isn’t a Titan from NVIDIA anymore – in fact, they did consolidate that next-gen hardware into the RTX Lineup – so, AMD claims that pitting Radeon AI GPU against the RTX 5080 16GB, shows drastic potentials in terms of performance on a good few Large AI Models, namely Phi 3.5 MoE Q4, Mistral Small 3.1, DeepSeek R1 and more. On top of that, with the scalability you need in such environments, AMD is providing exactly that with the Multi-GPU PCIe 5 platform.
Pricing for the AMD Radeon AI PRO R9700 will vary according to the partners. But at the time of reporting, we have no pricing on it. Stay tuned for more updates.
For more tech, gaming, and lifestyle content, make sure to follow us on Facebook, Instagram and Twitter. Better yet, stay tuned to The Adventures of Vesper.
