The upscaling methods in direct comparison
Before we get straight to the point, I would like to say a few words of a general nature. I’m only comparing the processes against each other today, and only in the Spider-Man Remastered game. The topic of DLSS 3.0 and frame generation does not fit in today. And for good reason. I spent over an hour swinging through the city, running around and jumping off buildings. At over 200 FPS with DLSS and FG on, I could not perceive any artifacts that can result from frame generation. PG279QM thanks!
This has a technical reason: At 240 FPS / 240 Hz, the image is displayed for a ridiculous 4.16 ms. If you want to recognize something, you have to record the whole thing, and this is where it becomes unbelievable. Den some other sensationalists probably did just that. If you record a game at 60 FPS that runs at 240 FPS, you suddenly see every frame for 16.6 ms and if you then watch it at only 25% speed, you will of course also see the “fake” frame that comes from DLSS 3.0 as an intermediate frame. If you watch 240 FPS on a 60 Hz monitor, you will have a similar problem.
For this reason, you have to be damn careful what you put out into the world. I think this is borderline – almost unbelievable. For my part, I simply refrain from doing so. I know very well that FG is not perfect. The less FPS you have natively or even with FG on, the more likely it is that you will actually notice the fake frames. MS Flight Simulator in UHD and CP 2077 UHD on Psycho Settings would be real examples here. That is my opinion and I stand by it – even if FSR 3.0 comes with interframe calculation. If this works as well as DLSS 3.0, I won’t badmouth it for no reason! That was my impression when I watched some of the other articles and reviews on the launch of Frame Generation.
2160p native resolution
Here’s a reference of what the whole thing looks like natively:
DLSS vs. FSR vs. XeSS Quality @ 2160p
As nice as NVIDIA’s ICAT theme is, it is unfortunately also impractical. A few seconds of video in UHD equals 45 MB times three. With my article about the GeForce RTX 4090 FE, the CMS soon crashed and my internet connection was at its limit. For this reason, I use today unfortunately again YouTube, which is simply better to handle and already the server regarding the amount of data. Loading on mobile devices will also be easier for you readers.
DLSS has by far the most stable image, no AA flicker and can show the fine details best. XeSS is close to DLSS in terms of detail representation. FSR is still doing a decent job from an overall impression. In terms of displaying small and fine details, FSR cannot keep up with DLSS and even has to admit defeat to XeSS. The full drone of the individual videos can be found here: Click DLSS / Click FSR / Click XeSS
DLSS vs. FSR vs. XeSS Performance @ 2160p
Due to YouTube’s compression, it unfortunately doesn’t come across quite as well, but since the conditions are the same for everyone, that’s no big deal. There is actually no longer any doubt as to who is the true number one upscaler. DLSS performance outshines the competition. Take a look at the individual videos, too, then you’ll see it even better: Click DLSS / Click FSR / Click XeSS
I see that there are still doubters among you, for you I have something else: Because UHD can anyone…
1080p native resolution
DLSS vs. FSR vs. XeSS Quality @ 1080p
Here are the videos: Click DLSS / Click FSR / Click XeSS
Still in doubt or are there other opinions on who is the clear number one? Basically, we don’t even have to let the details speak, the first impression is enough. Here’s the last hard fact, even though it probably hurts some, but facts are facts – you can’t replace them with feelings.
DLSS vs. FSR vs. XeSS Performance @ 1080p
Here are the videos: Click DLSS / Click FSR / Click XeSS
The picture quality speaks a very clear language in the end: And the winner is – NVIDIA DLSS. The render resolution here was the same for all methods. So there are no excuses either. You can find the table here: Click!
The ladies and gentlemen from AMD and Intel simply have to do a better job. I think now even the last one has understood, because Spider-Man is not an isolated case either. End of the announcement!
Motion performance @ 1440p
In the end, I was surprised by all three methods, since none of them really made a negative name for itself with ghosting in the test.
Here you can see all three methods in the performance level with the target resolution 1440p. If you are interested in the quality level as well: click here
If you jump down from somewhere, then you also have to climb back up:
Here, slight ghosting occurs with XeSS. Well recognizable by the hands or the fingers. Here still the video in the level quality: click here!
I think that should be enough for today. Then I’ll come to the end… last page please.
- 1 - Einführung und Testsystem
- 2 - Cyberpunk 2077 @ 2160p
- 3 - Cyberpunk 2077 @ 1440p
- 4 - Cyberpunk 2077 @ 1080p
- 5 - A Plague Tale: Requiem @ 2160p
- 6 - A Plague Tale: Requiem @ 1440p
- 7 - A Plague Tale: Requiem @ 1080p
- 8 - Bright Memory: Infinite @ 2160p
- 9 - Bright Memory: Infinite @ 1440p
- 10 - Bright Memory: Infinite @ 1080p
- 11 - Spider-Man Remastered @ 2160p
- 12 - Spider-Man Remastered @ 1440p
- 13 - Spider-Man Remastered DLSS vs. FSR vs. XeSS
- 14 - Zusammenfassung und Fazit
87 Antworten
Kommentar
Lade neue Kommentare
Mitglied
Mitglied
1
Moderator
Urgestein
Veteran
Urgestein
Veteran
Moderator
Veteran
Moderator
Veteran
Veteran
Mitglied
Veteran
Veteran
Alle Kommentare lesen unter igor´sLAB Community →