FSR 2.0 Tested with Deathloop - A Huge Win for AMD
FSR 2.0 VS DLSS - Performance - Does DLSS run faster than FSR 2.0?
Published: 13th May 2022 | Source: OC3D Internal Testing | Price: |
FSR VS DLSS - Which has the most overhead?
Radeon's FSR 2.0 upscaler and Nvidia DLSS technology work in fundamentally different ways. FSR 2.0 is designed to be cross-platform and use generic GPU resources, while DLSS is designed to use the AI/Tensor performance of Nvidia's RTX series graphics cards. That explains the performance differences that we see below.
DLSS and FSR use the same input resolutions in the Performance, Balanced, and Quality profiles, but when FSR uses generic GPU hardware resources, DLSS uses a hardware block on Nvidia's RTX GPUs that are otherwise unused. Where AMD's FSR takes resources from other tasks, DLSS uses the untapped AI performance of Nvidia's RTX GPUs. The result is that FSR has a larger performance impact than DLSS, making DLSS faster.
While DLSS is faster, it is worth remembering that it only works on Nvidia RTX series GPUs. For non-Nvidia users, and users of older Nvidia GPUs, FSR 2.0 is the only game in town. If you can't use DLSS, it doesn't matter if it runs faster on Nvidia's RTX hardware.
Most Recent Comments
I'm not so sure. On plenty of visual media, 'sharpening' is often added but for me it reduces quality. Obviously so, since you are modifying the original source to create an illusion of clarity - but sharpening is not making things more clear. Nor is FSR. It replaces actual data with guesses and adds sharpening.
If the creator (the game studio) did not want the image to be super-sharp, are you improving it or degrading it when you 'enforce' sharpening? On the other hard, if the creator failed to get a image as sharp as they really wanted - one could argue you are now adding quality.
Sort of similar, you can turn up the treble and bass to music, and many will say it sounds better. But it is now a less correct representation of the original.
So how do we define "better"? Is it simply what people like? Or is it reproducing imagery as close to the creators intent?
It would be interesting to hear from the creators of Deathloop, what they think of the result. If they could say so without being afraid of AMD.Quote
|
Originally Posted by MiNo
As for 'better than the real thing' : Is "sharper" always 'better' ?
I'm not so sure. On plenty of visual media, 'sharpening' is often added but for me it reduces quality. Obviously so, since you are modifying the original source to create an illusion of clarity - but sharpening is not making things more clear. Nor is FSR. It replaces actual data with guesses and adds sharpening. If the creator (the game studio) did not want the image to be super-sharp, are you improving it or degrading it when you 'enforce' sharpening? On the other hard, if the creator failed to get a image as sharp as they really wanted - one could argue you are now adding quality. Sort of similar, you can turn up the treble and bass to music, and many will say it sounds better. But it is now a less correct representation of the original. So how do we define "better"? Is it simply what people like? Or is it reproducing imagery as close to the creators intent? It would be interesting to hear from the creators of Deathloop, what they think of the result. If they could say so without being afraid of AMD. |
As far as sharpness goes, FSR 2.0 has a sharpness slider that can be lowered to suit your preference. By default, the slider is set to max, and in some areas the game may benefit from dialling this back a little.
When it comes to looks, things are going to be very subjective in most cases. This is especially true for sharpening, as everyone has a different taste.Quote
CSS Valid