Sony revealed the technical details of PSSR 2.0 for the PS5 Pro. In the second generation of the upscale, the

Sony revealed the technical details of PSSR 2.0 for the PS5 Pro. In the second generation of the upscale, the company abandoned the approach where the neural network actually calculated the final color of each pixel, and switched to a hybrid architecture.

Now, the neural network is primarily responsible for image recognition, geometrical boundaries, and noise filtering, and HDR color mixing and final reconstruction is done by traditional GPUs.

Most of the neural network operations run at a resolution of 540p and below. Sony also replaced the previous output filter 3×3 with elliptical filtration based on the Gauss distribution.

Thanks to changes in architecture and quantization optimization, the model's training time has been reduced from about four months to one week.

However, contrary to rumors, PSSR 2.0 does not use transformers the system is still based on specialized swivel algorithms.

I don't know what you're saying. #game