Since the release of PS5 Pro in November 2024, Sony has continued to work on optimizing game image quality and resolution for this most powerful console. Although PSSR technology performs well in some games, other titles still have visual flaws and image issues that affect performance and picture stability.
To this end, Sony, under the leadership of system architect Mark Cerny, is working closely with AMD to integrate key technical elements of FSR4 into the PSSR of PS5 Pro. The latest patent filed with WIPO titled "Graphics Processing" reveals the technical direction of PSSR 2.0 for the first time, and the update is expected to be launched along with the development kit in the next few months.
The patent describes a new method designed to avoid frame rate drops and native resolution problems in PS5 Pro games. In short, developers may not need to lower the base resolution of their games to use PSSR in the future.
The core concept of this patent is real-time quantification. When the game encounters a large data load, the frame rate may briefly drop for a few seconds. The solution for PSSR 2.0 is not to reduce the native resolution, but to temporarily reduce the use of convolutional neural networks to stabilize performance.
CNN is a convolutional neural network, which is a machine learning system used by PSSR for image reconstruction and resolution improvement. When the frame rate drops, PSSR 2.0 will reduce CNN calls and bring performance back to normal levels without reducing the native resolution of the game.
According to the patent description, real-time quantization actively reduces the intensity of the neural network when the graphics processor load is too high, helping to smooth frame rate fluctuations, especially in fast-paced games.
This is still a technical concept at the moment and its actual performance in real games has yet to be proven. If it can achieve the desired results, Sony will hopefully solve one of the main problems observed by players on the PS5 Pro PSSR.