But it will be twice as fast as FSR 2
AMD has revealed more details about its upcoming FSR 3 supersampling technology.
The new technology will compete with DLSS 3, but if DLSS 2 and FSR 2 are technically very similar, then FSR 3 will not follow in the footsteps of DLSS 3 and will not build entire frames.
The new version of the technology will generate even more pixels in the frame using interpolated frames, which will allow the company to achieve a twofold increase in performance compared to FSR 2.
For those who are afraid of “fake” DLSS 3 frames
AMD said that FSR 3 will use Motion Vectors and AMD Fluid Motion to create interpolated frames. The new technology will improve motion estimation and will use additional internal information from FSR 2.
Frame interpolation is more complicated, as there are several problems:
1. We cannot rely on color fixing to correct the color of outdated samples.
2. Non-linear motion interpolation is difficult with motion vectors in 2D screen space, so we recommend entering at least 60 fps.
3. If the final frames are interpolated, then the user interface and all post-processing must also be interpolated.
However, there is good news too!
1. There is a high probability that there will be at least one sample for each interpolated pixel.
2. There is no feedback loop, as the interpolated frame will be shown only once – any interpolation artifacts will remain for only one frame.
But where AMD will again follow in the footsteps of a competitor is in reducing latency. In addition, FSR 3 seems to be very easy to integrate into games that already support FSR 2.
