AMD’s graphics division is in the throes of change. In fact, at IFA, Jack Huynh announced a few things about the company. The first was that the brand would be concentrating on more streamlined graphics card products. Next, we learned that the RDNA and CDNA architectures would be merged to simplify development. Now we learn that the FSR4 will run on artificial intelligence!
FSR4: like Intel and NVIDIA, it will be based on AI!
Until now, the various versions of FSR have not exploited artificial intelligence. Indeed, earlier versions exploited spatial and then temporary scaling techniques. Now, with the fourth generation, the technology will be based on AI frame generation.
The need behind all this is to extend the autonomy of portable devices, such as ROG Ally consoles. As Jack Huynh himself says, “I need to play Wukong for three hours, not 60 minutes”. One example given would be to block FSPs at 30 or 35, thus reducing hardware consumption and improving device autonomy.
Of course, there’s no need to panic: the portable sector is mentioned here, but this new version of FSR will be supported by graphics cards in due course. However, the wait shouldn’t be long now, as Jack Huynh announced that he had put his teams to work on it 9 to 12 months ago.
On the other hand, OC3D rightly wonders whether this announcement has anything to do with the announcement of the PS5 Pro, i.e. whether this upscaler is based on that of Sony’s console or vice versa. After all, it’s AMD!