- Radeon AI Pro R9700 was confirmed for the launch of July 23 with a 32 GB GDDR6 memory
- The lists of early partners show prices between $ 1244 and $ 1277 depending on the model
- The GPU is designed for demanding workloads such as AI, rendering and inference workflows of the local model
AMD’s Radeon AI Pro R9700, its fastest GPU to date, is now on sale, at a price of $ 1,244.
Although AMD did not confirm the exact prices when writing this BenchGive a good idea what to expect.
Ads show that Sapphire models at the price of $ 1,244 and $ 1,277, while the ASROCK version, available in pre-order on Tech-America is listed at $ 1267. This puts the probable median point at around $ 1,259.
Built for professionals
It is first indicated on Computex 2025, the R9700 is built on the architecture RDNA 4 and uses the GPU Navi 48. (Funny fact: its name is a nod to the 9700 pro original manufactured by ATI which outperformed Nvidia products in 2002 and helped to move the market. AMD then acquired a few years later.)
The card is delivered with 32 GB of GDDR6 memory, twice that of the Radeon RX 9070 XT, which shares the same nucleus but intended for players.
This memory works at 20,000 MHz on a 256 -bit interface, offering 640 GB / s of bandwidth, and could make a real difference for professional users working with large data sets, training models or juggling several pro applications.
By reducing memory exchanges to RAM, the R9700 aims to keep things smoother during demanding workflows.
The double-locity shape factor and the blowing style cooling target multi-GPU configurations in professional workstations.
The card offers up to 47.8 TFOP in performance with a single precision and up to 1531 tops in int4 operations. It contains 128 AI accelerators and supports most modern media codecs, including AV1, H.264 and H.265 for coding and decoding.
The card connects via PCIe 5.0 and has several output ports, including DisplayPort 2.1A and HDMI 2.1b.
Personalized models of people like Asrock, Asus, Gigabyte, Sapphire, XFX and Yeston will be available for purchase in the coming months.
These conceptions will include different cooling configurations or small adjustments, but the nucleus and the GPU memory will remain the same.
