When AMD initially announced that it was working on FidelityFX Super Resolution (FSR), an image magnification and enhancement algorithm, applicable to all Best graphics card And allow AMD to compete with Nvidia DLSS, It raises a lot of questions. Will it use machine learning, or some other upgrade technology? AMD clearly claims that the algorithm was built “fully in-house,” but People have seen the source code And determined that it is mainly based on the existing Lanczos resampling algorithm, which has existed for decades and has been used by Nvidia for its sharpening filter. Ah, the joy of open source software.
To be fair, AMD did more than just reuse Lanczos resampling directly. Specifically, FSR includes some optimizations to make it run faster, and some other filters to help eliminate any halo caused by sharpening. But, perhaps most importantly, AMD strongly supports the creation of an open source solution that game developers-or anyone-can integrate it into their applications. The idea behind upgrading and enhancing video content is not new, but sometimes it takes some elbow grease to get everyone on the same page.
A key point of FSR is that it should be integrated into the game, so it only applies to 3D rendering content, not to other content such as UI elements or text. Of course, nothing prevents people from using FSR to upgrade everything—in fact, there are several projects designed to do this—but some things are best presented in their original resolution. By optimizing and standardizing the FSR algorithm, AMD has managed to make at least two dozen games adopt the technology, as well as the Unreal Engine and the Unity Engine.
Ironically, Nvidia also adopted Lanczos upgrade and sharpening as a filter in its driver, which was first added many years ago. Does it allow developers to use Lanczos upgrades instead of time upgrades in the Pascal GPU generation-or in addition to time upgrades? Absolutely. But instead, it’s just a filter, and Nvidia uses its engineering efforts to create DLSS, an AI-driven upgrade and enhancement algorithm that is proprietary to Nvidia RTX GPUs. To be fair, DLSS 2.0 and later work very well, arguably better than Lanczos resampling.
But as we pointed out in the FSR test, there is a general algorithm that can be used any Modern GPUs—from Intel UHD 630 to Nvidia RTX 3090 to AMD RX 6900 XT—have many benefits.For example, if we look at the current Steam hardware survey (Focusing on DirectX 11 GPU), we found that in the recent survey, only 17.6% of PCs have RTX cards. This means that more than 80% of the game market cannot currently use DLSS. In contrast, every PC in the survey with a DirectX 11 or higher GPU should be able to use FSR.
As a former software developer, I can attest to the fact that when a new feature benefits 100% of the expected user base, it is much easier for management to approve the new feature, not just a small number of potential users. The needs of the majority are greater than the needs of the minority, or something similar.
Of course, the evidence lies in eating pudding. FSR pudding tastes almost as good as virgin pudding—maybe a little undermixed, but you will hardly notice, at least when using super-quality or premium profiles. When what we should really do is ask what has taken so long, let us not be too happy to congratulate AMD for creating something new and useful.
If one of the companies behind our modern graphics API—that is, Microsoft and DirectX or Khronos and Vulkan/OpenGL—integrated the Lanczos upgrade five or ten years ago and promoted it as a useful feature for game developers , We could have learned from the technology of those years. However, maybe people can use a slightly older GPU for a year or two before upgrading, which may not be considered in the best interests of AMD or Nvidia until now.
In any case, FSR is very useful, even if it is just a modification of the algorithm decades ago. Sometimes the best solution is the old solution. If you have an RTX card, DLSS is also great-sorry, GTX and AMD users. Hopefully we will see more widespread adoption of these two technologies in future games, because choice is almost always a good thing.