A few days ago, we learned that AMD is preparing a new FSR 2.0 technology, which will explain the details at the GDC 20222 Developer Conference, which has better performance and picture quality than the original technology, does not require specific AI acceleration, and even supports NVIDIA and Intel graphics cards. Now, here comes the further details.
FSR 2.0's algorithm will change from spatial scaling to a whole new generation of Tempora Upscaling, combined with optimized AA anti-aliasing, to provide comparable or even better picture quality than the native game screen.
Compared with FSR 1.0, the image quality is better in all quality grades and resolutions.
The most tempting thing is that FSR 2.0 does not rely on specific ML machine learning hardware, does not take the road of AI acceleration, so the hardware adaptation range is extremely wide, not only supporting its own products, but also supporting NVIDIA, Intel and other competitors.
Of course, whether the opponent uses it or not is another matter, and AMD did not say whether it will open source FSR 2.0.
In terms of performance, AMD gave an example, a game in 4K resolution, open the frame rate of 56FPS in the case of light chasing, open the performance mode of FSR 2.0, continue to turn on light chase, the frame rate is as high as 101FPS, almost doubled!
At the same time, there is better picture quality and higher performance, which can really be achieved, of course, depending on the optimization of game support.
FSR 2.0 will be officially released on March 17, a week before GDC, along with RSR (Radeon Super Resolution).
RSR can be seen as a simplified version of FSR, integrated at the driver level, with the advantage of supporting almost any game without optimization, limited to supporting only RX 5000/6000 series graphics cards, and the picture quality is naturally one notch lower than FSR.