Intel set to supercharge XeSS with a DLSS 3 rival that aims to tackle the worst thing about Nvidia’s frame generation
Intel is improving its frame rate boosting technology, XeSS, with a step forward in the form of ExtraSS, which promises to keep up with the latest developments with Nvidia's DLSS and AMD's FSR.
If Wccftech Spotted, at SIGGRAPH 2023 in Australia, Intel revealed that it is working on an extension of XeSS called ExtraSS, a frame generation feature that could rival the efforts introduced in DLSS 3 and FSR 3.
The main difference from Intel's take on frame generation – which literally refers to inserting extra frames into the game, to increase the frame rate and make it artificially smoother – uses frame extrapolation.
This is in contrast to AMD and Nvidia, which both use frame interpolation.
Intel explains: “We are introducing ExtraSS, a new framework that combines spatial supersampling and frame extrapolation to improve real-time rendering performance. By integrating these techniques, our approach achieves a balance between performance and quality, generating temporally stable and high-quality, high-resolution results.”
Analysis: Keeping up with the Joneses
This is a necessary step for Intel to keep pace with its GPU rivals. XeSS is a pretty impressive take on increasing frame rate in general, but with frame generation being the new king of upscaling, Team Blue seems to be falling behind if it doesn't look at its own implementation. Enter ExtraSS (we don't care about the name, by the way).
As mentioned, there is another approach for Intel that takes the extrapolation route, instead of interpolation. What does this mean? As mentioned, this is all about the way the artificially inserted frame is generated. With interpolation, it is produced from the data of the two frames on either side (before and after). However, when extrapolating, it is based on only one frame.
To summarize in an admittedly rough nutshell: Intel ExtraSS uses only one frame and therefore guesses more about the composition of the inserted frame, but the advantage is that there is no increase in input delay (which is caused when drawing of data from two frames so that the frame can be placed between them).
So Intel scores in reducing input lag – and that's where AMD and Nvidia's frame generation technologies have come in for the biggest criticism – but this is to some extent at the expense of the quality of the result (there's more guesswork involved) and the possibility of added artifacts and the like. However, it's clear that Intel will work to keep the latter to a minimum, and Team Blue will want to strike a tasty balance between performance and quality.
That said, whether extrapolating or interpolating, frame generation is still essentially the same technology, so Intel's ExtraSS won't be everything That different. It's just that Team Blue will clearly be looking to gain a slight edge over its rivals in terms of input lag. (Note that AMD and Nvidia also have their own anti-lag features to counteract any harmful effects of frame generation – at least in supported scenarios).
Ultimately, the proof will be in the pudding, and we're eager to see how Intel can work its magic on the frame generation front when ExtraSS actually comes to fruition – although we don't know when that will be.
It may take some time for ExtraSS to come to fruition, especially if FSR 3 is to be believed, which took about a year to arrive (and even more time to iron out wrinkles, anyway). Meanwhile, Nvidia has made the move to DLSS 3.5 with a new twist on the upscaling technology: ray reconstruction – and it's a serious step forward. There's no doubt who's leading the FPS boost race, but AMD has also been making good progress with FSR 3 lately.
Through Tom's hardware