AMD to Update Radeon GPUs to FidelityFX Super Resolution 2.0 Soon, Radeon Super Resolution Available Today

What to look forward to: Today, AMD has put together a bunch of announcements related to GPU software. The biggest one is the launch of a new version of FidelityFX Super Resolution. FSR 2.0 will be a temporal scaling solution that is very different from FSR 1.0, a spatial solution. Temporal FSR 2.0 brings it more in line with technologies such as Nvidia’s DLSS and Intel’s forthcoming XeSS, which use temporal data to improve the upscaling process and simply provide the algorithm with more data points.

Switching from temporal to spatial, FSR 2.0 should provide a much higher visual quality than FSR 1.0. FSR 1.0 was useful at resolutions like 4K, especially when using Ultra Quality mode, where it could get close to (but not match) DLSS at times.

But to put it simply, the image quality that DLSS was able to deliver at lower resolutions and lower quality levels was much better – think 1080p or when using Performance mode. There are also elements that DLSS handled better, such as foliage, and FSR could struggle when the game’s built-in implementation of TAA was poor, as we saw in Marvel’s Avengers.

Many of these areas of image quality need to be improved when switching from spatial to temporal. AMD claims that FSR 2.0 should deliver “similar to or better than native image quality” and that it includes “optimized anti-aliasing”, suggesting that FSR 2.0 will replace in-game anti-aliasing on activation, as DLSS does.

…to say that something gives “better than native” image quality has become a buzzword these days.

However, to say that something gives “better than native” image quality has become a buzzword these days. We’ve seen Nvidia claim this countless times, and it’s not always the case.

Does AMD mean better than native when looking at static scenes? In move? At lower resolutions? We’ll have to look into this as we get more information and hands-on time with the technology, but at the very least they expect FSR 2.0 to deliver better image quality than FSR 1.0 at all resolutions and presets.

Another key aspect of FSR 2.0 is that it does not require special machine learning (or AI) hardware to run. In fact, when we asked AMD, they said that no AI was used in the creation of the algorithm, so in other words FSR 2.0 is not AI based like DLSS is (kind of). The reality is that DLSS in its current form only uses a generic AI algorithm to improve the scaling pipeline – the same algorithm applies to all games. It is possible to create a generic non-AI temporal upscaling algorithm like Epic Games did in Unreal Engine, it’s just a question of how good it is and we don’t have any idea about it without testing yet. This.

AMD will talk more about how FSR 2.0 works at the GDC presentation next week.

However, the advantage of not using AI is that AMD can create a solution that works with a wide range of products. AMD states that FSR 2.0 will run on AMD GPUs and competing GPUs, however, they have not gone into detail on which GPUs will be supported or which instruction sets are required.

For example, Intel XeSS is AI-accelerated and supports two pipelines, one using the XMX acceleration found in Intel Arc GPUs and the other using DP4a instructions. DP4a is widely supported on current generation hardware, but not as much on older cards like the Polaris series (RX 580 etc.).

If AMD’s solution can use an even more widely supported instruction set than DP4a, that would be a huge bonus for owners of older GPUs, but we’ll need to hear more at GDC to see the specific requirements for FSR 2.0 and how to do it. working. Hopefully this will remain an open source solution and we very much expect it to stay that way.

AMD was ready to show the image quality for FSR 2.0, but only in one scene in Deathloop, and only as a static image. There are a bunch of caveats here, but let’s take a closer look.

AMD provided native 4K resolution, FSR 1.0 at 4K using quality mode, and FSR 2.0 at 4K using quality mode. FSR 2.0 clearly delivers better image quality than FSR 1.0, even when using the Quality mode based on this single, perhaps ideal, pattern.

The sign in the center of the screen is sharper with FSR 2.0, while the brickwork on the left has finer detail, closer in quality to what you see in the original image. FSR 2.0 also has clear signs of temporal upscaling and “better than original” image quality in fine detail.

The advantage that FSR 2.0 brings to this scene becomes even clearer when looking at AMD’s performance mode comparison. FSR 1.0 looks pretty bad here, especially for small details. Look at how these railings and the window in the upper left corner have been processed. FSR 2.0 is very jagged and clearly lacks detail as it cannot reconstruct these areas well from a single 1080p frame. But FSR 2.0 looks very similar to native except for some obvious differences in shadows in these shots, not sure if this is related to FSR 2.0 or not, again all we have are these images.

Now let’s talk about warnings. The first is obvious in that it shows a sample of one game. The quality of FSR 1.0 varied considerably depending on the title. Some games didn’t play well with the technology, others were decent. It’s the same with DLSS, where in some games it pretty much gives a better image than the original, while in others it can reduce and blur details. Deathloop may be one of the best examples for FSR 2.0, and AMD chose not to directly compare it to DLSS in this game, which is available.

Second, we’re looking at a static image. It’s not even a video of a static scene, it’s literally a screenshot. Static, non-moving scenes captured as screenshots are extremely favorable for temporal upscaling because they hide any issues in motion. The word “temporal” in technology refers to how data is collected to scale, temporarily, over time. And this is how artifacts occur, if the scene changes very little from one frame to the next, this is the ideal dataset to use for temporal scaling. But if the image changes, it becomes much more difficult for temporal scaling to maintain the original image quality.

We have seen this many times before with DLSS and other technologies. DLSS has been prone to severe ghosting in several games. This has been fixed to some extent in newer versions, but may appear in FSR 2.0. Like flickering and moiré effects for fine details, this can be quite noticeable when using lower quality zoom modes in motion. Nothing we’ve shown today shows how FSR 2.0 will handle motion, and this is arguably the most important factor in image quality.

As expected, AMD also offers a significant performance boost, although again we only have one example that shows Deathloop going from 53fps at native 4K resolution to 101fps in FSR 2.0 performance mode.

We’re very excited to see how FSR 2.0 compares to DLSS and XeSS, but it’s all about performance and image quality in the end. I’m also curious to see how a non-AI based approach can handle this problem, especially if that path allows it to run on a lot more hardware. So many questions and so few answers… yet.

Moving on to other AMD announcements, let’s now talk about the Radeon Super Resolution, which was first announced at CES 2022. The RSR will become available today as part of the new AMD Software Adrenalin Edition package, so you can download it and try it out for yourself if you have you have an RX 5000 series GPU or later.

Radeon Super Resolution is a driver-based implementation of FSR similar to Nvidia Image Upscaling. While FSR is an implementation in the game itself and is applied before the final effects and UI are rendered, RSR is applied to the entire look of the game and does not require developer integration. The trade-off here is that RSR supports more games (basically any game) but with lower image quality since it will scale things like the user interface that shouldn’t ideally scale.

The AMD RSR implementation has some advantages that Nvidia Image Upscaling does not. Most importantly, you can apply it per game in the driver settings, oddly enough Nvidia doesn’t let you do that, which makes it clumsy to apply only to certain games.

However, using RSR remains a multi-step process: first you need to enable RSR in the driver, either globally or for the game you want to scale, then separately you need to lower the output resolution of the game in the game itself. In an ideal world, this would be a magical one-click solution like FSR, but it would be very difficult to get it to work with a wide range of games, so AMD tried to come up with the next best thing.

These are AMD’s main announcements today. The latest Radeon drivers also include a few other useful features such as image sharpening support for media playback, updates for AMD Link, etc., but RSR support today and FSR 2.0 coming soon are big.

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button