It feels like it was just yesterday when Sony hardware architect Mark Cerny was For the first time we tease the “PS4 successor” from Sony and its “advanced ray tracing capabilities” based on new AMD chips. Now that we almost five full years into the PS5 erait's time for Sony and AMD to start teasing new chips that will form the basis of what Cerny calls “the console of the future in a few years.”
IN short nine-minute video posted ThursdayCerny sat down with Jack Huynh, senior vice president and general manager of AMD's Computing and Graphics Division, to talk about “Project Amethyst,” a joint project between the two companies that also teased back in July. And while Project Amethyst's hardware currently only exists in simulation form, Cerny said the “results are very promising” for a project that is still in its “early stages.”
ML, less problems?
Project Amethyst is focused on going beyond traditional rasterization methods, which don't scale well when you try to “brute force it through,” Huynh said in the video. Instead, the new architecture focuses on making the machine learning-based neural networks at its heart more efficient. AMD FSR upscaling technology And Similar PSSR system from Sony.
From the same source. Two branches. One vision.
My good friend and fellow player @cerny and I was recently reflecting on our shared journey, symbolized by these two pieces of amethyst chipped from the same stone.
The Amethyst project is a joint development between @PlayStation And… pic.twitter.com/De9HWV3Ub2
— Jack Huynh (@JackMHuynh) July 1, 2025
While this type of scaling currently allows GPUs to deliver 4K graphics in real time, Cerny said that “the nature of the GPU is fighting us here,” requiring the computation to be broken down into subtasks that will be processed in a somewhat inefficient parallel process by separate GPU compute units.
To get around this problem, Project Amethyst uses “neural arrays” that allow computing units to communicate and process problems like a “unidirectional artificial intelligence engine,” Cerny said. While the entire GPU won't be connected this way, connecting smaller sets of compute units will allow for more scalable shader engines that can “process more of the screen at a time,” Cerny said. This means Project Amethyst will allow “more and more of what you see on the screen… to be affected or improved by machine learning,” Huynh added.
