AMD Reveals Details About FidelityFX Super Resolution 2.0 aka FSR2

Image Credit:

At GDC 2022, AMD unveiled its FidelityFX Super Resolution 2.0 algorithm, which we’ll refer to as FSR2. GDC is geared towards game developers; FSR and FSR2 are open-source algorithms. Now, you can imagine what kind of performance improvements they can offer. What’s more, AMD has provided details of how everything works, as well as the decisions and optimizations made. It also discussed some of the advantages and drawbacks of FSR 1.0, as well as how things would change with FSR2.

Just like any new update, the FSR2 is designed to give considerable quality improvements over FSR1 on a high level. Moreover, the FSR2 is a brand-new algorithm that isn’t based on previous FSR work, integrating the new API will necessitate rewriting some code. 

AMD reveals that FSR2 will also be more demanding on hardware since it focuses on improving image quality. Anyways, the new algorithm is an open API, which is delineated to function on AMD’s competitors’ GPUs. Surprisingly, the new algorithm is compatible with previous-generation GPUs. However, the performance advantages might not be the same on lower-end hardware.

Also Read: AMD Rolls Out Ryzen 5000 CPU Support for 300 Series Motherboards.

Details about FSR 1.0

FSR 1.0 was a high-performance spatial upscaling method that was simple to include in games. However, there was a fair share of flaws in that algorithm. For instance, it demands a high-quality anti-aliased source image. In other words, if a game didn’t already have anti-aliasing; it had to add it before it could utilize FSR.

Furthermore, the upscaling performance was directly dependent upon AA. As a result, if low-quality AA is there, it will produce lower upscaled output. In addition to that, it also struggled with lesser resolutions. Since, there was not a lot of information in the source frame which caused shimmering, poor edge reconstruction, and other abnormalities.

What’s new in FSR2

FSR2 brings a temporal upscaling approach, and it has far more complicated input requirements than FSR. For example, FSR2 considers everything from the scene color, scene depth buffer, and motion vector buffer instead of just a frame. This is comparable to the DLSS method but FS2 requires no machine learning hardware.

If we go into details about integration, FSR2 takes an aliased image as input and produces anti-aliased results as output.

What’s more, FSR2 looks relatively simple to add into games that currently use DLSS or other temporal upscaling techniques. FSR2 also supports dynamic resolution scaling (DRS), which makes it a viable alternative to other dynamic upscaling systems.

Furthermore, AMD has also tweaked with various upscaling options. There were four presets in FSR1 before- Ultra Quality, Quality, Balanced, and Performance. However, FSR2 will feature Quality, Balanced, Performance, and an optional Ultra Performance mode. Since the company has harmonized its mode names with DLSS.

Further, it appears as if the input resolutions are unaltered, so Quality at 4K still upscales from 2560×1440 to 3840×2160, Balanced at 1.7X, Performance at 2X, and Ultra Efficiency at 3X. It is worth noting that FSR2 is now in “beta” mode, with more improvements planned.

Also Read: AMD Unveils Six New Models As a Part of Spring 2022 Ryzen Desktop Processor Update.


Leave a Reply

Your comments will appear once it is approved by a moderator.


Galaxy S22 US pricing

Samsung S22 Ultra Users Facing GPS Issue

Roku Brings Better Personalization with Roku OS 11

Roku Brings Better Personalization with Roku OS 11