The latest advancements in frame generation techniques from GPU manufacturers, such as NVIDIA’s DLSS 3 and AMD’s FSR 3, have caught our attention. However, on the Intel side, there have been recent developments as well. At SIGGRAPH Asia 2023 in Sydney, Intel introduced their ExtraSS technique.
ExtraSS is an extension of Intel’s previous frame scaling technique, Intel XeSS. What sets ExtraSS apart from DLSS 3/FSR 3 is that Intel utilizes an extrapolation method for frame generation, whereas other companies use interpolation.
Interpolation involves using multiple frames as samples to create a new frame. This process combines the previous and subsequent frames to generate an intermediate frame (see illustration). The advantage of this method is that the resulting frames have high quality. However, the downside is the increased processing time, leading to latency issues. To address this, NVIDIA has introduced additional techniques like Reflex to reduce latency.
On the other hand, Intel’s Extrapolate technique utilizes a specific frame as a reference. The benefit of this method is lower latency. However, the disadvantage is that the quality of the generated frames is lower since only one input frame is used. What Intel presents is an Extrapolate frame generation technique that offers improved quality (though not equivalent to interpolation) while still maintaining low latency.
Currently, ExtraSS is still in the research phase and has been presented at graphics conferences. Intel has not announced when it will be implemented in real-world applications.
Illustration: Comparison of Interpolation (left) and Extrapolation (right) frame generation techniques – NVIDIA
Source: Wccftech
TLDR: GPU manufacturers like NVIDIA and AMD have showcased advanced frame generation techniques like DLSS 3 and FSR 3. Intel has also made strides with their ExtraSS technique, which uses extrapolation instead of interpolation. While interpolation offers higher quality frames, it introduces latency issues, mitigated by techniques like Reflex. Intel’s Extrapolate technique provides lower latency but sacrifices some quality. ExtraSS is currently in the research phase, with no announced implementation plans.
Leave a Comment