AMD FSR 2.0: Everything you need to know about this new image upscaling technology

Last March, AMD presented FSR 2.0, an image rescaling technology that, as its name suggests, is an update to the current FSR 1.0. At the time we already told you some of its most important keys, and we saw a graphic comparison that pitted both technologies both against each other and with the native resolution.

Overall, the conclusion we were able to draw from that comparison was that FSR 2.0 is indeed well above FSR 1.0, and that it seems to have reached such a good level that it has nothing to envy to the native resolution when viewed. used in its highest quality mode. Still, I think it’s recommended. that we keep expectations in check, and that we wait to see more realistic tests, and in motion, before drawing any kind of conclusions.

Although the article that we published at the time was very complete, in the end we left some things in our “virtual” inkwell so as not to saturate the article too much. Certain questions that were left up in the air are quite important, and for this reason I wanted to share with you this new article, where we are going to review everything you need to know about FSR 2.0 adopting a question-and-answer model. If after reading it you have any questions, you can leave it in the comments and we will help you solve it.

What is FSR 2.0 and how does it work?

Is a image upscaling technology which is capable of generating frames at a higher resolution than the one it uses as input. So, for example, you can generate a 1080p image from 720p samples. Unlike FSR 1.0, it is not limited to working spatially, but rather also contains temporary elements. This means that it takes into account the elements present in frames prior to the one we want to generate, something that helps improve the final result.

To further improve the final result, this technology also applies edge smoothing, which reduces the effect of «aliasing» or «saw teeth» which is usually caused by reducing the total number of pixels in the scene. In the first images we’ve seen, FSR 2.0 does a good job, and it’s clear that the combination of temporary and special elements has a lot to do with this.

What modes can FSR 2.0 use?


AMD has confirmed that FSR 2.0 will have a total of four different modes, but these will be different than what we saw in FSR 1.0, as ultra quality mode is dropped. These are the available modes and their most important keys:

  • Quality mode: delivers image quality equal to or better than native, while maintaining 67% of the pixel count of the target resolution. Thus, for example, to reach 4K, it starts from 2.56 x 1,440 pixels, and to reach 1080p, it starts from 1,280 x 720 pixels.
  • Balanced mode: it strikes a balance between image quality and performance gain. It might not reach the level of a native image. It reproduces 59% of the total pixels of the target resolution, so to reach 4K it starts from 2,259 x 1,270 pixels, and to reach 1080p it has a base of 1,129 x 635 pixels.
  • Performance mode: this mode prioritizes performance, and achieves “native-like” image quality, according to AMD. It preserves 50% of the pixels of the target resolution, so to reach 4K it starts from 1,920 x 1,080 pixels, and to reach 1080p it starts from 960 x 540 pixels.
  • Ultra performance mode: it is a mode designed to offer the highest possible performance, while maintaining an image quality “representative of the native”. It only keeps 33% of the pixels of the target resolution, so be clear that it will only be a good idea if we want to play in 4K. To rescale to that resolution, start at 1,280 x 720 pixels.

What graphics cards will support FSR 2.0?

At first, it seems that all the ones that are with FSR 1.0. However, AMD has given a list of recommended models depending on the resolution we want to use. This recommendation has an explanation, and that is that FSR 2.0 has a slight impact on performance, since it is ultimately a rescaling algorithm that must be executed and applied, and its use on lower-power GPUs could end up having consequences. undesirable.

These are the recommendations that AMD has given:

  • 1080p: Radeon RX 590, RX 6500 XT, or GeForce GTX 1070-GTX 16.
  • 1440p: Radeon RX Vega 56, Radeon RX 5600, or GeForce GTX 1080, RTX 2060 or better.
  • 4K: Radeon RX 5700, RX 6700 XT, or GeForce RTX 2070, RTX 3070 or better.

What games will be compatible with FSR 2.0?


We don’t have a definitive list yet, but AMD has confirmed that this technology it will be very easy to apply in those games that are already compatible with NVIDIA DLSS, and also in those that are based on the Unreal Engine 4 or Unreal Engine 5 graphics engines, thanks to the available plugin. All in all, we can be sure that a good amount of titles will end up using FSR 2.0

In the coming months, we will share with you a list of compatible games that will be updated over time, but only when it is really necessary, that is, when a considerable number of available titles accumulate. It is important that you do not confuse it with AMD RSR, since this rescaling is present at the driver level, while FSR 2.0 has to be integrated directly into games.

Will I need dedicated hardware to enjoy FSR 2.0?

No, AMD has kept the same approach that we saw with FSR 1.0 technology, which means that FSR 2.0 does not use any kind of dedicated hardware to work, which NVIDIA DLSS does since, as our regular readers will know, this image rebuilding and upscaling technology relies on tensor cores.

You can use FSR 2.0 on any compatible graphics card, although keep in mind that AMD has warned that if we have a model that is below the recommended ones, the results could vary, that is, they may not be as optimal as those. It is a somewhat ambiguous statement, but I think they mean that performance gain might not be that great.

What difference will there be between FSR 2.0, DLSS and XeSS?


The main difference is that DLSS and XeSS they use artificial intelligence. In the case of NVIDIA, that is the reason why it runs through the tensor cores, while in the case of Intel the XeSS will have two modes, one accelerated by hardware and another by software that will be compatible with a multitude of cards. graphics, which puts it in a level of more direct competition with the FSR 2.0.

We still don’t know how FSR 2.0 will actually position itself against DLSS and XeSS, but in theory the use of artificial intelligence should give a clear advantage to NVIDIA and Intel in performance and image quality. FSR 2.0 would have the advantage, over DLSS, of being multi-platform and not requiring dedicated hardware, but in this sense it will compete directly with Intel’s XeSS, which will not require specific hardware either, as we have already said.

What performance improvement can I expect?

I still can’t give you an exact figure, but I can give you some data that will help you as reference:

  • The performance improvement you will get with FSR 2.0 will be greater the higher the graphics load. This means that there will be more performance gain from downscaling from 1080p to 4K than from 720p to 1080p.
  • At the same time, the performance improvement will be greater the lower the base pixel count. This means that you will get a higher performance increase with the balanced mode than with the quality mode, and with the lower ones.

However, keep in mind that when moving in very low resolutions, not only may we no longer notice a performance increase, but this can end up being negative, that is, we could lose performance. This is because we reach a level where GPU usage drops and the CPU ends up being a bottleneck, either due to lack of power or the game’s own limitations. Thus, for example, it does not make sense to use the ultra performance mode with 1080p resolution, but it does with 4K.

Related Articles

Leave a Reply

Your email address will not be published.