Sony's new patent will enable 8K gaming on PS5 similar to Nvidia's DLSS

      A new patent from Sony Interactive Entertainment shows that Sony is developing a machine learning technology for real-time image enhancement for PS5 and VR. This machine learning process is somewhat similar to Nvidia's Deep Learning Super Sampling (DLSS) in that it can increase the resolution of games without sacrificing performance.

　　Nvidia's DLSS allows computers to run games better by rendering the original image at a lower resolution, making it easier for the GPU to process, and using AI to upscale the image to run at a higher resolution. This means that visually intensive games can be run on a 4K screen or even higher settings, but FPS will not be affected because the game is actually running at 1080p, but the look and feel are very similar to the actual 4K resolution.

　　Sony's patent, filed in April and published just last month, outlines a method where machine learning can create a "mask" of a specific image to hide or fill in gaps left by missing or corrupted data. Essentially, this allows for real-time image enhancement without sacrificing performance, as the technique is less taxing on the hardware.

　　The patent mentions VR several times, and this technology is essential to providing higher quality images, which is what a premium VR headset would need. Powering a VR headset requires a lot of processing power, and it’s worth noting that the PSVR headset has a low resolution.

　　The new patent describes machine learning technology as using "less computing power" and "not sacrificing image quality where it's most needed to maintain user comfort and immersion." This could mean the next generation of PSVR headsets could run VR games with better visual quality and performance.

　　As we progress through the 2020s, we could see more 8K screens enter the mainstream market, and this technology could also be used to upscale images to 8K.

　　The author of the patent is Sony senior principal engineer Andy Bigos, who writes on LinkedIn: "Working on the application of machine learning and neural rendering to real-time graphics on PS5."

　　“Recently, I have been pioneering the use of neural rendering to augment traditional rendering methods, focusing on using implicit neural representations, and how to make them run efficiently. This includes creating custom high-performance perfs using compute shaders. This work is targeted at PS5.