On October 28, 2025, news emerged that AMD's latest-generation professional graphics card, the Radeon AI PRO R9700, officially made its debut in the retail market on October 27. Earlier, in May, some of its specifications were tentatively revealed on paper. Then, in July, it was formally launched, offering pre-installation options for OEM (Original Equipment Manufacturer) and SI (System Integrator) manufacturers.
The Radeon AI PRO R9700 is built around a full-fledged Navi 48 chip. It comes packed with 4096 stream processors, 64 ray-tracing accelerators, and 128 AI accelerators. In terms of memory, it boasts an impressive 32GB of GDDR6 ECC memory. When it comes to power consumption, the entire card draws 300W and is powered by a single 16-pin power connector.
In the realm of computational performance, this graphics card truly shines. Its FP32 vector performance can reach a remarkable 47.8 TFLOPS (TeraFLOPS, a unit measuring floating-point operations per second). For FP16 matrix operations, it can achieve up to 191 TFLOPS. When dealing with FP8 matrix sparsity operations, it can reach an astonishing 766 TFLOPS. And for INT8 matrix sparsity operations, it can deliver 766 TOPS (Tera Operations Per Second).
The reference design of this graphics card is provided merely as a guide for manufacturers. The non-reference versions available on the market either adhere to the reference design or are independently crafted. Regardless of the design approach, all of them are equipped with a single turbofan for effective cooling. As for the price, the official starting price is set at 10,999 yuan (equivalent to $1,299 in North America).
