When DLSS launched in 2019, it was an unexpected hit. Nvidia made the switch from GTX to RTX to highlight how important ray tracing will be to GeForce GPUs, but it actually turned out to be a surprising killer feature we didn’t think we wanted is his DLSS. Of course, every time Nvidia makes something better, AMD has to do the same. And in 2021, the company finally launched his DLSS competitor, FSR. Here’s everything you need to know about FSR, what it is, whether it does anything good.
FSR: A lightweight challenger to DLSS
Source: AMD
FidelityFX Super Resolution (FSR) is software that, like DLSS, can improve image quality and indirectly increase frame rate. First released in 2021 for games such as: Anno 1800 and dota 2. However, one of the main differences between FSR and DLSS is that AMD’s Radeon chips don’t come with AI hardware, so the former doesn’t use his AI hardware. For the end-user, not using AI doesn’t make much sense, but it does have a big impact on FSR and how it responds to his DLSS.
Just as AMD has copied many branding schemes from its competitors, the various versions of FSR are basically similar to the three versions of DLSS. FSR 1.0 is very similar to the original DLSS in that the image quality improvement is very small. However, FSR 2 is the definitive latest version and, like DLSS 2, offers significantly better image quality than its predecessor. FSR 3 hasn’t been released yet, but it’s been confirmed to implement DLSS 3-like frame generation (although AMD calls it frame interpolation), and this isn’t about improving image quality, but more. It is intended to improve performance by creating more frames.
One of the benefits of not needing AI is that FSR works on almost any GPU from any vendor and is officially supported down to cards from 2016 onwards, so whether you own an older model or have the latest and greatest GPUs. In contrast, DLSS 1 and 2 only work with RTX branded cards and DLSS 3 is further limited to RTX 40. It’s not entirely clear at this time if FSR 3 will receive the same level of support as FSR 1/2.
How FSR resolution upscaling and frame interpolation works and why it’s not perfect
Source: AMD
Unlike DLSS, FSR relies on fairly traditional image rendering techniques such as anti-aliasing to improve image quality, but combines it with algorithms to increase fidelity a bit more. Even though FSR is advertised as capable of improving performance, it does not. Rather than using FSR to make it look good at one resolution, AMD uses it to make lower resolutions look like higher resolutions, and this is how they improve performance. 720p cannot be 1080p with FSR enabled actual It’s 1080p, but the resolution is lower and it looks closer to 1080p.
The basics are: FSR 1 first spatially upscales the actual rendered frame, then sharpens. It’s nothing special and relied on the game to provide good anti-aliasing, so the actual frames the GPU renders have the best shots upscaled properly. FSR 2, on the other hand, is much more complex, replacing anti-aliasing while adding a number of other features such as depth and motion vector buffers. Basically, FSR 2 makes a significant change to the rendered frames over FSR 1, making his FSR 2 even better in improving image quality.
As I said before, FSR 3 hasn’t been released yet so we don’t know how it will perform, but AMD has revealed that it’s pretty much the same as DLSS 3. While running a rendered frame through the image enhancement algorithms of FSR 2, FSR 3 takes the difference between two different processed frames and creates a frame in between. New algorithms are used to make these frames so accurate that it looks like they were actually rendered by the GPU. The performance improvement should be around 50%, which is not bad.
Rather than using FSR to make it look good at one resolution, AMD uses it to make lower resolutions look like higher resolutions, and this is how they improve performance.
Of course, FSR shares almost all of the problems inherent in DLSS. First, FSR 1 and 2 do nothing when a CPU bottleneck occurs. If your CPU is bottlenecking because it’s too slow, too old, or your game is poorly coded, it means that lowering settings won’t really increase your framerate. increase. This is because these settings are almost exclusively for the GPU, not for offloading. CPU. Resolution is a purely GPU-dependent graphics option, and if you have a CPU bottleneck, lowering the resolution, like FSR 1/2, will not improve framerate. This is only realistically solvable with very good CPUs.
Regarding FSR 3, it seems to have the same problem as DLSS 3. FSR 3 had to replicate literally everything that would normally appear in a finished frame, including UI elements, and DLSS 3 has proven to have a very hard time replicating it. Realize your UI without introducing any visual artifacts. Even if the DLSS 3 struggles with its flashy AI core, it’s hard to imagine the FSR 3’s fairing being much better. Another issue is latency. Frame 2 (the interpolated/generated frame) is displayed after frame 3 has been rendered, as it takes 2 frames to create a frame in between. This means your game will react a few milliseconds slower than usual. Visually you’ll see a lot of frames, but it’s as if you’re playing at a lower framerate instead of a higher one.
To be clear, FSR 3 has not yet been released at the time of this writing, so these issues are only theoretical, but will almost certainly be observed once FSR 3 is released. prize. Latency is high because it is literally impossible to reliably create a frame with only a single reference unless FSR 3 can somehow predict the future. AMD can’t avoid latency issues, but at least it’s possible (although highly unlikely) that the FSR 3 could handle his UI elements better than his DLSS 3.
FSR is still not as good as DLSS, but it has one big advantage
The FSR 2 is also good, but reviewers agree that Nvidia’s technology is generally better. It’s featured in more games, has higher visual quality, and is the only thing that really matters to the end user. However, FSR has two major advantages that DLSS does not.
The first is compatibility. FSR 1 and 2 work with almost any GPU. Nvidia has been making RTX cards since 2018, but many are still using older GTX branded cards from the 10 and 16 series. These users can’t use DLSS because their card doesn’t have his Tensor Cores, but they can use FSR 1 and 2. It might seem ironic that Nvidia (and Intel) users would benefit as much as FSR 1 and 2 owners. It’s an AMD card, but it brings a lot of utility to the FSR.
The FSR adoption rate is also very important. DLSS gave him a head start in nearly three years over his FSR, but both technologies are used in similar amounts of games. At the time of this writing, it appears that about 300 or so games have some form of DLSS, but as of December 2022, 230 games had FSR, so perhaps 250 games now have DLSS. I can guess that it has FSR 1/2. This is by no means a bad level of support for FSR, and it also means that FSR is being added to more games sooner than his DLSS.
That’s probably because it’s easier to implement in games, at least according to AMD. A game that lacks all the technology FSR 2 needs to work can take up to 4 weeks to be able to run FSR 2, but adding FSR 2 to a game that already has DLSS 2 takes may take as little as 3 days. In contrast, the developer should work directly with Nvidia for his DLSS support. FSR is open source and it’s up to developers to integrate it into their games. Of course, the obvious problem is that games with DLSS 2 are limited. So if FSR momentum hasn’t slowed yet, it could definitely slow down.
