After testing AMD FSR 2.0, I'm almost ready to ditch DLSS

For a few years, Nvidia’s Deep Learning Supersampling (DLSS) was the reason to buy an RTX GPU. AMD’s FidelityFX Super Resolution 2.0 (FSR 2.0) update will change that. I took it out to shoot Deathloop, it’s the first game to get support and this upgraded tech is the DLSS killer I’ve been waiting for.

Of course, there’s a significant difference between the two from a technical standpoint, but what I’m focusing on here is image quality and performance.

FSR 2.0 is essentially a branded version of Transient Super Resolution (TSR), which first appeared in Ghostwire Tokyo March. Although I need to watch more games with it to fully count the DLSS, Deathloop It’s a hugely impressive performer, both in terms of picture quality and performance.

Performance FSR 2.0

Deathloop have FSR 1.0, FSR 2.0 and DLSS, so I can do some side-by-side comparisons of performance and image quality. I’ve looked at the Quality and Performance modes with each upgrade method, they offer 1.5x and 2x scaling, respectively.

Starting with Quality mode, it’s no surprise that FSR 1.0 leads the way with a 66% improvement in my average frame rate over native 4K. DLSS came in second with a 55.7% increase, while FSR 2.0 came in second with a 48.2% increase. FSR 2.0, certainly behind, but all three upscaling modes boosted my frame rates dramatically, even at their most modest quality modes.

Performance mode is where the FSR 1.0’s picture quality begins to degrade. And to be fair, this is where DLSS also starts to see some limitations. Even when scaling to twice the original resolution, the three upscaling methods are just as relevant as they were with the Quality preset. However, DLSS and FSR 2.0 are much tighter, delivering gains of 96.3% and 92.9%, respectively.

Performance benchmark for FSR 2.0 in Deathloop.

It might seem like a loss for FSR 2.0, but as you can see in my chart I ran these tests with an Nvidia RTX 3090. FSR 2.0 is not locked to my graphics cards. AMD, unlike DLSS, only works with RTX 20- and 30-series GPUs. DLSS provides trivial performance is better, but the difference of 4.9% (maximum) is hard to justify a purchase given how expensive Nvidia’s graphics cards are, given current conditions. And it’s even harder to justify when you can use FSR 2.0 with a six-year-old GTX 1070.

FSR 1.0 leads the way, but the key thing to remember is image quality. Our FidelityFX Super Resolution review showed how this upscaling method breaks down in active quality modes. FSR 2.0 may not offer much performance, but it looks much better.

Image quality FSR 2.0

FSR 1.0, FSR 2.0, and DLSS all boost performance – it’s no surprise that they display fewer pixels. The real comparison is in image quality, which is what causes so much of the gap between DLSS and FSR 1.0. FSR 2.0 changes that with a much smarter approach to transient data-associated hypersampling.

First for image quality is Quality Mode. The pictures below are one of the better areas in Deathloop to see the image quality due to all the fine details in the background. In FSR 1.0, you can see the “Open Mind” sign in the pixelated jumble, as well as balustrades on the base in the lower left.

Compare quality mode for FSR 2.0. — FSR 1.0 (from left), FSR 2.0 and DLSS.

FSR 2.0 and DLSS is where things get interesting. They look almost identical. The main difference is that it seems that FSR 2.0 leans more towards sharpness. There’s a bit more clarity on the larger textures (see buildings to the right) and more detail on Cole’s jacket. Therefore, sharpness also pronounced aliasing on the wire towards the top of the image.

While playing, can’t see the difference between FSR 2.0 and DLSS. FSR 2.0 enhances sharpness slightly, but both methods of supersampling look very close to native resolution. The shocking thing is that AMD achieved this image quality without training some AI models or charging extra for dedicated Tensor cores.

Quality mode is one thing, but I wanted to push FSR 2.0 as far as I could. I also took some screenshots with Performance mode and I zoomed in 267% to really see the pixels. However, I cannot find a meaningful difference between FSR 2.0 and DLSS.

Compare FSR 2.0 in Deathloop, zoom in. — FSR 1.0 (from left), FSR 2.0 and DLSS.

This zoom level shows how terrible FSR 1.0 really is, but again, DLSS and FSR 2.0 look almost identical. I think FSR 2.0 actually looks a bit better. It doesn’t show aliasing around the floor lamp like the DLSS does, and it still maintains small details like the label on the bottle towards the right. Keep in mind how much I zoomed in here: If you’re trying to figure out the difference here, you definitely won’t see a difference while playing.

FSR 2.0 is the new champion

Deathloop's hero shoots enemies with two shotguns.

From the moment I saw TSR in Ghostwire Tokyo, I know FSR 2.0 will impress. However, I didn’t know it would make this much of an impression. DLSS used to be a reason to spend more on Nvidia GPUs, but even with the small performance differences, FSR 2.0 has rendered that feature obsolete.

DLSS is still technically better, so if you have access to it you should use it. However, FSR 2.0 means gamers don’t have to choose anymore. The only downside to FSR 2.0 at this point is that it’s not available in more games. Hopefully this promises to show up in Deathloop change that.