AMD has been hard at work for the past few years, and it has done a stellar job of catching up with its competition. While it has Intel beat for now in the CPU space, it still has some distance to go before it can bounce back to match Nvidia in the graphics department. It certainly is making progress there though, and the latest milestone in that direction is AMD AMD FidelityFX Super Resolution (FSR).
At the recently held Computex 2021 event in Taipei, AMD took the wraps off of its brand new GPU technology. It wasn’t new hardware, but rather a software solution that will rival Nvidia’s Deep Learning Super Sampling. Called AMD FidelityFX Super Resolution, it’s a way to get some extra boost in your gaming performance. Let’s take a deeper look at it.
Also read: What is FreeSync? AMD’s display synchronization technology explained
What is AMD FidelityFX Super Resolution and how does it work?
Simply put, AMD FidelityFX Super Resolution is a supersampling upscaling technology that promises to give a boost in the framerates in games, when enabled. At its core, it’s an anti-aliasing technology that takes color samples of the pixels at different instances. These values are averaged, and this is done by sampling images at a much higher resolution than the actual images. What this means is that the end result will be a more detailed image.
We have seen information surrounding AMD FidelityFX Super Resolution for a while, While AMD itself hasn’t given us a super-detailed explanation of how FSR works, we did get a patent filing that showed up days before the announcement. AMD’s implementation uses both linear and non-linear upscaling networks. This basically means that it aims to retain as much data as possible, to ensure the upscaled image looks as detailed as it can.
AMD FidelityFX Super Resolution takes a low-res image and puts it through the two upscaling networks. It picks out different features of the images through each of the two pathways. It then puts all the data together to create a pixel grid and expands it to make a higher resolution image, which is cleaned up and outputted.
AMD FSR has four presets that AMD showed off in the demo — Ultra Quality, Quality, Balanced, and Performance. Higher FPS modes come at the compromise of quality. This is because AMD FSR renders at the lowest resolution on Performance mode, and scales it up. Here’s a look at the render resolutions and scale factors at 4K and 1440p resolutions.
For example, in Godfall running on an AMD Radeon RX 6800 XT, AMD demonstrated a 4K native frame rate of 59 FPS. With AMD FSR running at Ultra Quality preset, the frame rate jumped to 87 FPS. Reviews have noted that the higher FPS modes show a noticeable lack in detail, due to the lower input resolution which means the source itself lacks detail which AMD FSR cannot itself populate.
Also read: AMD vs Nvidia – what’s the best add-in GPU for you?
FidelityFX Super Resolution vs Nvidia DLSS
AMD FidelityFX Super Resolution is AMD’s answer to Nvidia Deep Learning Super Sampling, aka DLSS. There are some stark differences between the two technologies, even though they seem like equivalents on paper. While both are upscaling technologies that use supersampling, both have fundamentally different approaches to image processing.
AMD FidelityFX Super Resolution is more or less a run-of-the-mill spatial upscaling technology, which has existed for a while. It does take a new approach, but it is fundamentally similar to those before it. On the other hand, the DL i.e. Deep Learning, in DLSS means that it is doing a lot of artificial intelligence and machine learning-based image corrections.
Nvidia DLSS uses a neural network to reconstruct images at a higher quality on the go. The result is a much better image output than what the traditional supersampling can provide. AMD FSR cannot match the level of detail machine-learning-based models are capable of achieving.
Additionally, being machine-learning-based, Nvidia DLSS required per-game training for version 1.0, which is why adapting the games to use DLSS has been a bit slow. On the other hand, AMD FSR announcement states that it “does not rely on history buffers or motion vectors. Neither does it require any per-game training,” which will make the implementation easier for developers.
The output AMD FSR gives you arguably looks worse than Nvidia’s first DLSS attempt. DLSS visuals are nearly matching the native output in terms of quality these days. With FSR, you will likely get the higher frame rates that AMD showed off. However, as you can see in the demo, and as many folks have pointed out already, there is quite a bit of visible blurring happening, which means AMD FSR isn’t quite matching DLSS.
Also read: What is Nvidia DLSS? Nvidia Deep Learning Super Sampling explained
AMD FidelityFX Super Resolution supported GPUs, games, and availability
This is where things get interesting with AMD FSR. Nvidia DLSS only supports Nvidia’s RTX lineup of GPUs. On the other hand, AMD FSR is open source. It supports RDNA2, RDNA, and Vega GPUs, as well as Nvidia GPUs using Pascal, Ampere, and Turing architectures. This includes AMD Radeon RX 6000, RX 5000, RX 500, and RX 400 series GPUs, and Nvidia GTX 10-series, as well as all RTX GPUs. IGN has also confirmed that Xbox Series X and S will support AMD FSR since the consoles use RDNA2-based GPUs.
AMD actually demonstrated the technology using Nvidia’s GeForce GTX 1060 GPU showing a 41% improvement in Godfall using the Epic preset. Make no mistake though, AMD has made it clear that it will be up to Nvidia to work on the implementation. AMD Radeon CVP & GM Scott Herkelman tweeted about the same.
Nvidia has not commented yet, but Intel has shown interest in supporting it with its new Xe GPUs. Intel’s architecture, graphics and software division chief Raja Koduri, who was previously an AMD executive, tweeted about Intel’s interest in supporting this open platform.
We saw a similar situation in G-Sync vs FreeSync, where Nvidia ended up supporting FreeSync under its G-Sync Compatible branding. The difference here is that AMD FSR isn’t a full alternative to Nvidia DLSS. AMD could and should make a version of FSR that leverages machine learning at least on Radeon hardware, to give better performance, if it truly wants to match DLSS.
See also: AMD GPU guide: All AMD GPUs explained, and the best AMD GPU for you
Which games does AMD FidelityFX Super Resolution support?
As far as supported games are concerned, upon announcement, AMD said that 10 game studios and engines were on board. The demo had only shown Godfall running with FSR. The company had also published a survey page asking users which games they would like to see FSR support.
AMD FSR has now launched, and AMD has announced that more than 40 developers plan on supporting FSR officially.
To begin with, FSR is available in seven games: Anno 1800, Evil Genius 2, Godfall, Kingshunt, 22 Racing Series, Terminator: Resistance, and The Riftbreaker. Additional games will support AMD FSR by the end of 2021. These titles include Far Cry 6, Resident Evil: Village, DoTA 2, and more.
Stay tuned for more on AMD FSR, PC graphics, and other relevant PC tech. In the meantime, check out these articles.