Deathloop is the first game to support AMD FSR 2.0. A comparison of AMD’s downsampling with predecessor FSR 1.0 and Nvidia’s DLSS: FidelityFX Super Resolution 2.0 shows much better, FSR 1.0 stands a chance. The test details the size of the differences and whether they are also sufficient for Nvidia’s DLSS 2.0.
FSR 1.0 vs. FSR 2.0 vs. DLSS 2.0
A year ago, AMD introduced FSR 1.0 (test), the number one competitor to Nvidia’s DLSS, whose approach without so-called temporary components also had its advantages. But overall, FSR 1.0 didn’t come close to DLSS. AMD was also aware of this from the start, and it wasn’t without reason that the new technology was born under the name “FidelityFX Super Resolution 1.0”. It was clear from the start: a second copy would follow.
With the FSR 2.0 announced in March 2022, after countless rumors, it was official that the second iteration would face DLSS 2.0 head on. Although still without machine learning in the background, but with a temporary solution that includes information from previous frames in shorthand, FSR 2.0 is following in the footsteps of DLSS 2.0. It can now be seen if the results can be linked to DLSS in the Deathloop.
FSR 2.0 premiere via patch in Deathloop
Because it’s a bit surprising that a patch for the Arkane Deathloop (test) comes out today and adds FSR 2.0 to the already supported technologies FSR 1.0 and DLSS. ComputerBase had a chance to take a first look at it a couple of days ago and will show in the following sections whether the qualitative leap is as big as hoped.
Following the latest update in the graphics roster, Deathloop also offers FSR 2.0 in Quality, Balanced, and Performance levels as a way to upgrade in addition to FSR 1.0 and DLSS 2.0 – developer Arkane does his job without the optional “Ultra Performance” option.
Currently, editors have only used the ‘Quality’ and ‘Performance’ levels across all three technologies. In FSR 1.0, on the other hand, “Balanced” and “Ultra Quality” were omitted due to time limitations. A target resolution of 3840 x 2160 will play a key role in testing, as this is where the extra performance is most needed. But 1,920 x 1,080 and 2,560 x 1,440 were also taken into account.
This is AMD FSR 2.0: the technology in detail
FidelityFX Super Resolution in the original 1.0 is a classic “spatial scale”. Spatial Promoter analyzes each submitted frame/image and attempts to improve the image quality based on this information. Unlike pure scaling, an attempt is made to identify the elements and objects in the image and then improve them in a targeted manner. The image is then optionally sharpened with FidelityFX CAS if the developer wants it.
FSR 2.0 uses a time component
On the other hand, FSR 2.0 is a temporary shorthand, like Nvidia’s DLSS 2.0. This means that the database relies not only on the current frame, but also on additional data from previous frames to improve image quality. Using the additional information, the image quality can also be improved in terms of image reconstruction and stabilization, which is not possible with a spatial overlay.
The lower the resolution of the display, the more advantages temporal downsampling can gain over the spatial variable from additional data from previous frames. But even with a high resolution, the advantages are obvious. This is why the temporal method can produce better graphics than at a higher original resolution, simply because there is more information available.
In addition, like DLSS 2.0, FSR 2.0 replaces the game’s own anti-aliasing and uses its own method to reduce annoying steps. FidelityFX CAS for clarity is still present, here the developer can also install the Optional Sharpness Controller.
FSR 2.0 and DLSS 2.0 are similar in many ways
AMD’s wording when describing FSR 2.0 is very similar to Nvidia’s when describing DLSS 2.0. Accordingly, FSR should be 2.0Picture quality equal or better than the original resolutionAccessible. But not only at this point are both technologies very similar in player presentation, the modes available are also nearly identical with Quality, Balanced, Performance and Ultra-Performance including internal display resolutions. “Ultra Performance” is only one option, the other three modes will always be available.
FSR 2.0 works without a neural network
Nvidia uses an automated process in DLSS 2.0 to combine information from previous frames, while AMD uses “It uses advanced algorithms that can detect relationships between different frames and resolutionsHowever, there is still a difference. With AMD, it should be up to the GPU alone to “understand” the image before upscaling, while with Nvidia, each supported game is first analyzed in a neural network. The ‘trained’ knowledge should further improve the outcome. It remains to be seen if AMD’s purely local approach works just as well as it does with a background neural network.
No neural network has the advantage that FSR 2.0, unlike DLSS, does not require matrix units (tensor cores) for acceleration (optional), and FSR 2.0 computations are performed on regular shaders. Additionally, AMD doesn’t tie the technology to its Radeon graphics cards, so FSR 2.0, like FSR 1.0, will run on any existing GPU — regardless of whether it’s from AMD, Nvidia, or Intel.
Additionally, at least in theory, an advantage is that FSR 2.0 potentially has fewer problems with classic DLSS problems such as ghosting or smearing. In the first example, the test shows if this is really the case – after all, errors can occur when adding data, even without AI. In addition, FSR 2.0 will be open source, although the source code is not yet available.
Lots of pros and cons vs. FSR 1.0.0 Update
In terms of quality, FSR 2.0 should be clearly better than FSR 1.0, at least in theory, but there will also be drawbacks. FSR 2.0 should initially only work with DirectX 12, and Vulkan support is expected to follow soon. However, there is currently no talk of more APIs. Additionally, FSR 2.0 will not be as executable as FSR 1.0 in Radeon Super Resolution (RSR) form in a game that does not already support FSR 2.0.
Because in order for the information from the various old frames to be properly taken into account, FSR 2.0 must be able to access the so-called motion vectors, which show the algorithm where the pixel “moved” to. Without this information, temporal shorthand cannot function. Aside from the vectors, FSR 2.0 should also have access to the depth buffer and color buffer – as a result, downsampling is integrated into the game rendering pipeline much earlier than FSR 1.0.
The developer effort is higher
Not for the player, but what is important for the developer is that the implementation of the second version like DLSS 2.0 takes longer than the first iteration. FSR 2.0 should be the easiest to integrate if the game already supports DLSS 2.0, since both technologies have nearly identical requirements. According to AMD, implementation should take three days, even if it is not clear how many employees have been estimated for this. In addition, there will be an FSR 2.0 plug-in for Unreal Engine 4 and Unreal Engine 5, which will significantly reduce integration times even without prior DLSS integration.
It should take less than two weeks if the game supports at least display resolution and resolution-independent TAA – because there are already important vectors for FSR 2.0. According to AMD, if TAA is not supported, about three weeks are planned. And if the game does not support TAA and resolution and display independent, the integration should take four weeks or more.