FSR 3.1 vs XeSS vs DLSS, Radeon vs GeForce, Spiderman and Miles Morales (is FSR Better on AMD?)
HTML-код
- Опубликовано: 31 июл 2024
- I look at FSR 3.1, DLSS 3.7, XeSS 1.3 and TAA in Marvel's Spiderman and Miles Morales with the focus being image quality. I also tackle the theory that FSR looks better on AMD in this video although it's not the focus. I'll make a separate video for that looking at multiple games.
FSR and XeSS footage used in VS comparisons was captured with an RX 7900XTX. DLSS and TAA footage was captured with an RTX 4090.
Music by: Karl Casey @WhiteBatAudio
00:00 Intro
00:58 Mirror Roof
04:23 Construction Site
05:39 Sunny Park
07:23 Cafe Cinematic
08:30 Phin's Base
09:30 MM Opening Cinematic
11:32 Courtyard
12:47 Ice Skater
15:44 Xmas Lights
17:03 Central Park
19:00 Final Thoughts
#fsr3 #spidermanmilesmorales #pcgaming 
Игры
Easily the best content I've seen in this subject!
Awesome work man.
The FSR on Nvidia vs FSR on AMD I think is finally an argument put to rest?
I think seeing the brighter lighting on the Nvidia gpu whilst running FSR showed the same characteristics were crossing over.
Was interesting however as Andrew (frogboy) has stated he feels he gets a better HDR image on his AMD gpu and I'm wondering if that's what he's pointing at, it actually looked quite good.
I also appreciate you point out native isn't always the gold standard.
I hope this content gets a lot views as you deserve credit for this 👏
I can't tell a difference between FSR on Nvidia or AMD and I primarily tackled it because I've been seeing this in my own comments section but also on Twitter and Reddit. Like I said I do plan on making a more dedicated video looking at multiple games on this topic. Theories are interesting to look at but my biggest gripe with this particular one is that I've seen some people in some comments sections of other peoples videos trying to invalidate their work demanding FSR be tested in AMD gpu's and it can be annoying because it takes a lot of time and effort to make these comparisons and do them well. For example it took me near twice as long to capture FSR footage twice, on my Nvidia and AMD gpu.
I like to sit back and apply logic to things though. We know how FSR works, it's open source, hardware agnostic. One aspect I could maybe see have an advantage on one brand over the other when comparing same class GPU is the amount of frames produced, since that relies on compute, but quality is a math algorithm that doesn't use dedicated hardware. Another thing I have noticed that there can be some assets that can load a bit differently across different brands of GPU in very rare cases.
The HDR thing I'm not really fit to test it in similar way. The best I can do is use my Samsung Galaxy S24 Ultra camera record my C1 oled screen in HDR and swap between the same game between my AMD and Nvidia PC. I have done this out of my own curiosity. They look the same, it's a standard. It's all up to the devs how well they implement HDR in their games. PC having lousy HDR narrative wasnt because of any particular brand of GPU but because of it being an afterthought with developers since its not until recently proper HDR monitors are becoming widely available. That's my thoughts on this.
@TerraWare Unfortunately I've witnessed the exact same on comments demanding the tests as invalidated..
As you say this content takes a lot of work so it's extremely frustrating!
Regarding what I said about HDR was just purely about the way FSR appeared to make the in game light bulbs etc glow much stronger..
@@Spratty-wb3is Since you have your own channel you will get some of this too. This is what I have come to realize. Firstly never try to reason with unreasonable people. some of these type of people have already made up their mind before even watching the content and even if you did the test their way, they can just say you are lying. You see wham I'm getting at? Next will be some who claim to get way more FPS than what you are getting with similar or weaker hardware. Lol. To be fair some can be misguided by having something like frame generation on and not knowing it or whatever.
I keep my eyes open and see what people say but if I sense someone is a fanboy type I dont even bother, it's pointless which happens with both brands btw. Its not just one. Can't please everyone just stay true to your principles and most people are reasonable and just share the same enthusiasms.
The HDR thing yeah I guess that makes sense. The color thing I'm not sure if you know this but if you are using Nvidia Shadowplay or AMD Relive to capture gameplay footage it will not pick up on custom color tweaks and change. It wont show up on the recording. The codec is already preconfigured. It's why sometimes if you put AMD and Nvidia footage side by side the AMD footage looks to have a greenish tint and Nvidia a reddish tint although in person playing they look the same. I have tested this. I made the colors black and white in the Nvidia and AMD driver and recorded gameplay footage it came out in color as it normally would in the recording.
@@TerraWare Funny as I had this very interaction already 🤣..
Someone asked me to test the FSR on Nvidia vs FSR on AMD which I believe he wanted to use in his own personal argument with a third party.
I agreed but at the same time told him it was pretty much pointless as no matter what I do my method will be challenged as incorrect 😂..
I'm not capable of the level in which you've achieved and I already directed him to your channel.
As I stated to him on a technical basis it wouldn't make sense, on a visual basis I notice no difference but I don't play a massive title of games.
And from what I can see if a difference exists it's that minor that it wouldn't effect the outcome enough in my opinion.
But yes I've noticed the colours on recording.
I'll be honest I don't mess with my colours on either system, I dunno if I've been lucky with good monitor calibration or I'm just not that fussy.
But both systems look very good to me..
@@Spratty-wb3is Exactly if there is a difference it's so insignificant that it's not worth obsessing about. That's one of the reasons why I didn't label which GPU was on which side on all my clips.
To test it properly though you have to use cinematics or gameplay on rails because take 16:43 for example if Spiderman is a bit further away on one side, that footage will be more flickery in comparison to the other, also moving the camera can cause flicker and shimmer and you can never align distance and camera movement with that precision.
I'll dig more into it on a separate video though looking at multiple games and come to a concrete conclusion and whatever I find I will show but I want to do my due diligence if that makes sense.
FSR 3.1 is more of the same unfortunately.
The issues that were present previously still are now, some to a lesser degree, and ghosting seems in fact to be worse.
I wasn't expecting it to close the gap with DLSS but it's pretty disappointing that even XeSS still remains a valid alternative to it.
We're basically in the same situation as before.
Yeah I was hoping for the flickering issues to be addressed personally as that's something I find quite distracting but at least now we can drop DLL updates on older games when AMD update it again. I've a feeling it will get better and better with time and in some games it does actually look pretty good. Like HFW and GoT when I looked at those games prior to the 3.1 update FSR super resolution looked really good and FSR frame generation was excellent.
@@TerraWare To me FSR 3.1 did look better on Horizon Forbidden West & Ratchet. Although i'm no an expert in looking for things but at the same time it doesn't really bother me as i'm focused on the gameplay so much.
DLSS is still ahead by a fair bit but it will depend on the person if that's such a big deal I suppose.
3.1 Frame generation is brilliant now & remember that had issues at the start, so i'm hopeful it gets better
PLEASE watch in 4K video quality for best image quality. This one took a while to put together and I had a bunch of footage and other areas left over that didn't make it into the video because it's already too long. The point of the video is to examine image quality with the latest upscales and point how they compare. It's purely academic and from an enthusiastic perspective. Let me know your thoughts and experience with these technologies. I will be digging into FSR 3.1 a bit more with the other Nixxes games and future games as they become updated. Thanks for watching and I appreciate it.
Best comparisation I’ve seen so far. Great work
Thanks. Glad you like it.
Out of all the games where FSR 3.1 was implemented, in Spiderman the issues stand out the most, I criticized Vex for not doing a proper comparison (And nitpicking) of 2.2 to 3.1, Horizon Forbidden West and Ghost Of Tsushima (Games that I play) look certainly GREAT to me with 3.1 in comparison to the previous version, ghosting is not that noticeable. By the way, great job in this comparison.
@silvio351 I haven't yet checked out the FSR 3.1 update on HFW and GoT but FSR 2 looked great, I did a video on GoT especially examining the generated frames VS DLSS 3 and it looked great.
Great job, @TerraWare and @Rhoberly were the only FSR videos with careful side-by-side viewport angles and backgrounds. This is specially important in ghosting issues. Only thing would be use if possible SMAA T2x instead of TAA
@@silvio351 The only issue I saw with Ghost of T was the ghosting around the character, apart from that there is an issue of flashing lights if you have recording enabled in AMD software or AMD instant replay. Turn that off and it fixes
I kinda always ignore SMAA
@@TerraWare Especially in Tsushima SMAA T2x has the best detail on particles in the first vilage looking from the mountain. Every other temporal AA/upscaler has severe issues with particles
Excellent work
Thanks!
Ghosting seems to be an issue on Ghost of Tsh as well with FSR 3.1. I gather that's something they will be working on to fix next because it's happening in Spider-man also.
Think Spider-man looks the worst version of FSR 3.1 out of all the games i've seen
Great video! I also wonder if one day you could take a look at comparing various DLSS presets AKA upscaling models using a tool such as DLSS Tweaks. These presets can be biased towards reducing ghosting or maximizing image stability for example, and there are many trade-offs in this regard. For example, I updated the DLSS Super Res DLL to version 3.7.0.0 in Robocop Rogue City and forced the game to utilize preset E, and it was able to completely eliminate some ghosting, at the expense of some image stability.
I would like to do that more yeah. I did that for Avatar Frontiers of Pandora where surprisingly dropping to the latest DLSS 2 from 3 got rid of the ghosting at release.
DLSS Tweaks is a great app. Aside from the ability to change presets, it also includes render scale option, so that base DLSS resolution values (50%, 59% and 67%) can be changed. I found out that 75% is a perfect value for those, who want the best possible visuals. At 1440p it looks identical to DLAA IMO, but still performs quite a bit better.
If only DLSS Tweaks could work with any game, which supports DLSS 2+. But it looks like sometimes it just does not work right. In Ratchet & Clank, for instance, it somehow produces severe ghosting artifacts.
Found this visually amazing I'm not a gamer great advances are being made 👍
It looks pretty good.
Great video. I think people say FSR looks better on AMD because they can't turn on DLSS to compare lol. On Nvidia cards you can simply use FSR and DLSS and see how much worse FSR looks in most cases, but I think Nvidia users never really need to use FSR ever these days. XESS 1.3 has been pretty good since they changed the resolution presets to give more FPS, so on AMD I might start using XESS 1.3 along with FSR 3.1 frame gen once both are in more games.
Haven't actually seen FSR 3.1 upscaling in person since I don't own any of these games, but the video comparisons I have seen seem to be all over the place in terms of quality, which is even more strange considering all the games are by the same developer. Maybe all are different engines in the back though since the original sony devs are all different.
Interesting, you haven't played any of the Sony games on PC? Must not be your type of game.
@@TerraWare Most of them I have played on console already before they were ported, but the PC ports are so late and so expensive. If they go cheap enough I'll probably double dip, but I think I might still end up playing most of them on PS5 with physical copies and sell them when I'm done. I do most of my PC gaming on game pass now and I doubt sony games would ever go on there.
God of War Ragnarok, Ratchet and Clank and Returnal are the main ones I haven't played yet which I want to. I don't mind waiting for the prices to drop way down though since they are all single player games.
On your last video where you presented fsr 3.1 in ratchet and clank I commented that it finally became a no-brainer option when set on quality at 4k. Now I think that AMD should really focus on AI enhancements in their upscaling technology. Only a few people(rx 5000 and older, and nv 10 series and older) still profit from fsr. RDNA2+ users might profit from those AI cores and NV RX and Intel Arc users have their own Upscalers, and the latter ones are better due to AI enhancement IMO
R&C was always the worst looking FSR out of the Nixxes ports with Spiderman being a close second imo. It's why I wanted to examine these and at 1440P. More recent implementations of FSR have been pretty good like HFW and GoT.
That said though I do believe AMD will take FSR in the machine learning phase. I also don't believe that they'd have to necessarily make it bound to their own hardware. They could do something similar to Intel or maybe use neural processing units that will be arriving in desktop CPU's soon. Sky's the limit.
I'm using HDR all the time on my 7900XTX and love it. Can't really stop using it now
Frog actually got me into it
😆 I'm really sensitive to really bright red colours and HDR makes it better to me on both desktop and screen
I use it too whenever it's a good implementation which lately we've been getting some loving with HDR.
@@TerraWare Yeah in Horizon Forbidden West it does really look great. Same as in Plague Tale requiem. Just annoyed me that there's no FSR in the game and the RT shadows seems bugged unless you have DLSS.
On their in game upscaler which honestly isn't very good at all, so much so I ended up modding FSR 3 into the game.
RT shadows however isn't really worth the performance hit for me, hardly looks any different.
You used to be able to mod FSR 2 mod in to overwrite DLSS but doesn't seem to work anymore.
FSR 3 mod does a good job but UI is unstable, I just turned subtitles off and then it's fine, definitely worth it to get better image quality
7:45 Look at the flickering in the hair with FSR. Both DLSS and FSR exhibit it, DLSS a bit less.
Yeah the flicker is always more stable on DLSS across the board.
BTW the "FSR works better on Radeon GPUs" rumor is nonsense.
I just don't ever use XESS because I want more FPS with FSR. I have tried it in a couple of games and just see it's less FPS
I've been wanting to look at the performance side of these upscalers. I do recall XeSS performance gains for equivalent settings being less than FSR.
@@TerraWare One game I remember was Witcher 3... I also thought even the image quality looked worse than FSR compared to XESS looking at Gerald's outfit. Think I needed to go ultra quality for XESS to clean it up better but was so long ago 😂
@@TerraWare Yes I seem to remember Ancient Gameplays doing some benchmarks and XESS FPS was lower most of the time compared to FSR
@@RJTHEGAME Look at Hardware Unboxed review. They measured the performance and XeSS somehow gives more fps on nVidia cards than on AMD cards and also DLSS 3.7 boosts performance more than FSR. Interesting for me because until FSR 2.2 the performance was similar on both DLSS and FSR.
@PeterPauls That was an interesting video. Especially the way he compared them based on FPS target. Not sure how much I liked it but he did compare same quality settings later in the video.
It was interesting nonetheless.
Fsr on amd and nvidia looks the same. Equally bad. The myth about open source code working differently on different cards is very funny. The quality of fsr depends on the fps, so the quality of fsr may differ for the worse on some gpu at low fps compared to high fps on another gpu)
@listoman9365 Right it doesn't really make sense how FSR could look better on different brands of GPU. The amount of frames generated and or latency, sure that can differ.
We are talking specifically about the quality of the FSR image at different fps. As the FPS increases, the FSR has fewer artifacts, ghosting, etc. I don't consider imputlag and fg)