Nothing like an opinion resource! Imagine finding a natural artesian well on the ground with just opinions flowing out of it. What an amazing resource! My opinion is DLSS looks like shit in motion. Either solution looks like garbage compared to native resolution however if you go from non-playable to playable FPS by using EITHER of these technologies, then who cares what it looks like. If you're using FSR or DLSS, you clearly don't care about image quality and just want frames. In that case, either solution works. Even Intels!
@@Cinnabuns2009 whats that? You need to eat, you have some bread, which you can eat to survive and the alternative is a pile of shit? Who cares ! Its not like either of them is Steak! Youre just after survivng so might as well eat either of them! Truely, fanboys cant be reasoned with. Insane amount of work done properly and fairly only for some random idiot to get butthurt over his favorit brand (be it amd/intel/ nvidia)
I still haven't forgotten about the time Nvidia didn't think Hardware Unboxed was deserving of their GPUs for review. This is a perfect example of why they shouldn't be left out. Great analysis with easy to understand results chart for potential buyers! I hope AMD improves on FSR. I am curious about the differences in FSR3 and the future they bring to a great feature regardless. Thanks Hardware Unboxed!
I've honestly not had an issue with fsr 2 in games. It is a worse looking upscaler than dlss2 but I only notice it when I compare side by side. In most games I've tried it in either looks good enough for me nowadays.
AMD will undoubtedly improve, but there's an upper limit to what they can achieve as DLSS is hardware boosted. Whatever improvements AMD makes Nvidia is likely to match and surpass due to this.
@@Freakmaster480 It's honestly the flickering instability that does it in for me. FSR just isn't great at keeping things relatively stable, it's super similar to UE4 TAA in it's earliest forms where stuff looks so jittery it annoys me.
@@Hotobu you know that two can play this game, right? AMD can also accelerate FSR with hardware in future generations of Radeon. My guess this is their plan but they launched behind so they needed an open source to get devs on board and also get much more user data and in the future they will release some form of FSR that's accelerated with hardware.
FSR doesn't have to beat DLSS. It just has to be in the same ballpark, but way more compatible to older cards and it is. People rocking a brand new-ish Nvidia card should probably opt for DLSS, but FSR works on way more cards overall, including old Nvidia cards Nvidia themselves have abandoned. I mean, do tell me which upscaling technology looks better on a GTX 1080 - DLSS or FSR... :)
Well, older card compatibility will lose relevance in the future when all reasonable modern gaming Nvidia GPUs are RTX. But broader compatibility still makes it more practical to include. FSR image quality may also improve past DLSS now that DLSS source code has leaked.
Nivida is not helping themself with thier unnesscary limits. I do not expect Dlss on a ibm PC or something but the fact that dlss3 is limited to 40 series is an artificial limit
i kinda prefer nvidia AND amd pushing the boundries of new hardware, why should they keep creating tech and features that work for cards 10 years old, we have no right as people who own old ass hardware to expect the best and latest features... dont you see that AMD have to continuously compromise to make that work and the result is they are falling behind on features and quality... thank god they didnt listen to guys like you when getting RT on board and they went hardware solution over what would have been disastrous software RT.
I feel like these videos would greatly benefit from more screenshots and zoom ins on specific examples. Only the most obvious issues are visible through youtube compression
@@m_sedziwoj you missed my point. These comparisons are viewed via youtube and it's awful compression. Details that would usually be obvious on a local render on a 4k TV can be blurred like crazy in motion on RU-vid. Not to mention, some people might be viewing these videos on phones or lower resolution screens as well. If you've ever seen a video of confetti on RU-vid, you know how much detail can be lost due to compression.
Exactly this. In my opinion FSR almost always looked better through youtube compression because it loses so much detail it encodes better but in person DLSS performance looks like fsr quality in motion
Yes exactly. I always check FSR and DLSS in all games that support it on my C1 TV at 4K and the differences between the two technologies stand out a lot more than these videos
'Advantage for FSR' actually sounds like... advantage for Nvidia consumers. Low end Nvidia consumers benefit from FSR, mid and higher end Nvidia consumers benefit from FSR OR the superior DLSS.
@@SpookyKG nah fam, it benefits anyone who has a compatible GPU be it NVIDIA, AMD or Intel. I have a rtx3080 laptop and I'd use either to be honest (even XeSS for all I care) if I need the technology for a specific game. Basically, either you get more options, or it's your only option, either way it's better, so there's no need to be a fanboy for any of these technologies. However, it's important to mention, AMD should be complimented for keeping it open source, AND cross-brand compatible.
Its funny how you say superior when every youtube channel is using Nvidia gpus to test both fsr and dlss. They should be testing amd software (fsr) with amd hardware (7900xtx) but seems every youtube channel is a shill for Nvidia for whatever reason
I still consider it impressive that FSR is slowly becoming a sidegrade to DLSS since there's no hardware techniques on the mix. Sure, DLSS is better for most people, but I strongly support solutions that are open for everyone.
@@NamTran-xc2ip That's sadly true. It's a really hard choice for consumers. Take a technology that could become obsolete or take a currently inferior tech that could become the norm? Only time will tell.
@@Dantido Think of it this way. Nvidia worked on DLSS, why should others, especially competitors benefit from their work when they spend time and money on it? They are a business, not your friend, not a charity just like AMD. If I create a product to sell, I wouldn't want others to also be able to sell it, unless with some fees or obligations. I bet my ass if somehow AMD has DLSS and Nvidia doesn't, they wouldn't make it open source.
@@NamTran-xc2ipI love remembering when amd made 3000 series threadripper its own socket and promising that “yeah it’ll have newer cpus in the future, don’t worry!”
Both FSR and DLSS are great technologies that would further help game optimization though I get the feeling recently they skip the optimization process all together now and just add FSR and DLSS and call it a day. A shame.
those technologies should have never seen the light of the day. calling them "great" shows that gamers just go along with everything. it will only get worse from here. and hardware unboxed's videos regarding this issue aren't helping either
This is an insane amount of testing, and the community is thankful for it. Shout out to Lost Judgment as well! So happy to see an RGG game mentioned by you guys!!
If you think about, that FSR has broad compatibility even with Nvidia cards it´s a miracle that it can somewhat compete with Nvidia´s hardwarespecific DLSS For me personally FSR is more than good enough (ok maybe not in Forespoken)
Biggest issue with these upscaling solutions i have is that the devs never implement Ultra Quality mode that renders the game at 3/4 of native resolution, even tho it's listed in the specs for both DLSS 2 and FSR 2. The 4K quality mode in my opinion is just lacking in general image clarity due to the render resolution being only 2/3 of native. Having only 2 options where you're either rendering the game at 66.6% or at 100% with absolutely nothing in between is just stupid. It's honestly insanely annoying that in almost every game there is no render resolution slider instead of these very limiting and nonsensical presets.
I assume it's because the algorithms are trained specifically for those resolutions, and perhaps the overhead of DLSS 2/ FSR 2 means that above 66.6% you wouldn't actually get any peformance gains, or if you did, they wouldn't be as great as those at 66.6% and the visual quality increase would not be worth it. Just speculation. You could use DSR (dynamic super resolution) in the Nvidia control panel to increase the resolution above your monitor's output resolution and then apply DLSS 2 to that in order to achieve your desired effect. Someone commented above that they do this with AMD's VSR (virtual super resolution, DSR equivalent) and achieve better results that way. Not sure about DSR though because classically it has been considered worse than VSR.
@@naturesown4489 DSR is kinda iffy in terms of upscaling but DLDSR on the other hand is pretty good(though limited by upscaling resolution options), idk if any comparison exists between it & VSR though.
I honestly think both look fine enough. Yes DLSS does have a better temporal stability, but FSR works on just about anything since 2016 whereas DLSS is RTX exclusive. Also the DLAA implementation of DLSS is more fascinating to me than the upscaling method. It is hands down the best temporal anti-aliasing method I have ever seen. Default TAA just looks like vaseline is smeared on the screen.
The big advantage of DLSS is that it is easily customizable. You can change DLSS version in any game, that supports it, by simply replacing DLSS file (which is located in a game folder). Ghosting issue can be easily fixed. For example, in Dead Space Remake stock DLSS version can be replaced with 2.5.1 and it eliminates ghosting completely. Same goes for Metro Exodus Enhanced Edition. Thus it makes DLSS even more superior to FSR. Plus there are plenty of DLSS modes for those games, which do not support it natively. I know there are many FSR modes too. But again even modded DLSS still usually looks better than native FSR, like it is in case of Resident Evil 4 Remake.
Yeah that choice between too soft or too sharp is something I've run into with DLSS before. It feels like it's a result of the game applying its own sharpening on top of DLSS, and in many cases that can't be disabled.
@@FatherLamb Yeah DLSS has that too, but some games have another separate sharpening filter that can't be disabled which is what seems to cause DLSS' oversharpened look.
@@FatherLamb Some games, like TLOU Part 1, have built-in sharpening that cannot be disabled without a mod or hex edit of the game executable. It makes the game look very ugly and over-sharpened. Note that I am NOT talking about the DLSS/FSR2 sharpening slider, I am talking about a BUILT-IN sharpening filter that the game uses.
DLSS version 2.5.1 is your best friend then. it has the sharpness completely deactivated, this means there also is no unsharpness like if you set the normal version you got on 0%.
Its cause FSR isnt using machine learning and devs can only do so much image reconstruction by hand with FSR2. FSR is legitimately only even passable at 4K quality also, dropping below that in anyway and it falls apart hard. People talk about how it saves old cards but it really doesnt and youre way better off just dropping settings. TSR looks about just as good as FSR and sometimes better, and XeSS IS also a lot better imo when actually using a Intel card. FSR is literally just a boxcheck to muddy waters vv Nvidia, rather than actually trying to compete they cut corners and tell you theyre just as good then try to charge you the same price till the market corrects for not being able to compete and drops prices for AMD. The shitty aliasing is the biggest indicator your playing at a lower res and FSR on 1440p and 1080p legit looks like im playing a PS3 game at 720p or something. And yet people who clearly never used DLSS say FSR looks just as good as DLSS which is just the biggest cap
@@Angel7black AMD needs to to follow suit and turn FSR 2 into AI upscaling and they'll probably do that in the future any way. They will probably turn it into a more advanced XeSS that works better on AMD cards, but also on other cards but at the quality level it has now for non AMD cards pre 7000 series.
DLSS 2.5.1 is still my favorite version by far. There is nothing to fiddle around with and this weird (over)sharpness is just deactivated. Absolutely recommend it for any game.
@@dontpokethebear3893 Why do you need DLSS swapper? I just drag and drop the DLSS 2.5.1 DLL file into the game folder and it overwrites the previous version. No need for some software program. I'm not arguing with you btw. I just don't understand why DLSS swapper is being used by some.
Great to see so much testing! Either way, on quality modes with the performance uplift, I didn't see either would be unpreferable. I'm simply not going to notice or be bothered by some shimmering or ghosting in the middle of gameplay. But what I would notice is my game dipping below target fps.
I am very sensitive to that kind of stuff and in the Resident Evil 4, I just turned off FSR2 and anti aliasing entirely at 4k. Looks way better than FSR2 quality mode. The difference is night and day to me. It is blurry, shimmery and makes me question whether my glasses are too weak. Even with adaptive sharpening manually added it just looks bad. It destroyes the textures.
@@LeegallyBliindLOL tbh in re4 remake I honestly couldn't tell between FSR on and off except on some of the herb items, so I just kept it on to allow me to max out everything else + RT
@@LeegallyBliindLOL yeah fsr was unusable to me in that game, I just played at 1440p native and took the performance hit, thankfully I have a 3080 so it wasn't that big of a deal but yeah i wish they had dlss in the game
The visuals around the characters head are perminantly blurry crap.. I can't unsee it once I've picked up on it LOL. It's one step forwards 2 steps backwards. The whole point is we're aiming for high refresh rates to remove ghosting + screen tear. But now people pay a premium artifact ghost simuatlor to reach "high frames" with a reduction in response time. It's freaking madness.
Yeah, the majority of people can't tell what's the difference between 1440p and 4K in a blind test, let alone some blurriness or ghosting at 200% zoom.
Welp, I spent pretty much 30 minutes looking at both screen mostly having no idea what the difference was. Except few that pops out at your face, I was basically Pam saying, "Its the same picture" Nice video though, Tim! I got a GTX so more FSR stuff is really appreciated.
@@spoonherr Yeah YT compression doesn't help and you don't know the type of monitors and resolutions people are watching these videos. Having tested FSR and DLSS myself on my C1 LG TV for example the differences are noticeable between the two.
@@Relex_92 I'm watching it on a 1080p monitor, maybe that's why. Though personally, I think it would bode well for me haha, I won't be replacing my 1080Ti anytime soon.
Fantastic side by side comparison. Not really a surprising result, but seeing that overview with DLSS and all the plus plus plusses is quite brutal :D Smashing work yet again, Tim!
Techtesters is another great channel if you like HWU. Nada often has great insights and a different perspective than Tim and Steve which helps to have a broader picture of things 😊
This only confirms what I was "feeling". Usually I go with 4k quality on dlss, I do spot things from time to time, but it's never too bad to the point that it breaks the inmersion. Hopefully FSR improves in the future, but right now the flickers drive me out from the game.
Don’t forget you can use DLSS Tweaks to trade in performance increases for higher quality. The base render resolution ratios can be changed to whatever you want and DLAA can be achieved at a 1.0 ratio.
I don't see how it would improve much without an equivalent to Nvidia's tensor cores. If AMD wants to compete in anything else than price, they need to really up their game.
I also favor DLSS except in games where the ghosting is rather obnoxious, like in Cyberpunk. In those situations, I personally will take FSR's shimmering and inferior fine-detail reconstruction, at least when using 4K quality.
@@timmyp6297 Yeah that looks like absolute crap sorry. Nvidia 3000 series barely lasted a year lol only the 3090/3090 ti is even usable anymore. I just sold my 3070 ti, nvidia is a joke with the vram. Unless you have a 4090 your nvidia card wont last a year, and is already outdated. 1440p is a wal mart budget resolution and with dlss you are basically playing 720p in 2023 LOL, too bad you nvidia users cant game in 4k without low textures lol
Things have really come a long way since 10 or 15 years ago! I know it's important to compare the current tech options to get good output for your money. So, thanks very much for the effort and the time put into this! But to my eyes, it all looks really breathtaking, no matter what settings you are using in the tests. The last gaming PC I bought had a mid-tier Core 2 Quad CPU and a Radeon HD 4000 series GPU. I was happy if games basically ran without issues at or often below 1080p. The eye candy you get in games these days... The sheer pixel counts! The effects! Real-time hardware encoding for live streaming all on the same GPU... AI post-processing features??! Whew! All that is to say, I feel like all these options look tremendous, and I'd feel pretty spoiled if I plunked down the cash for one of these machines today and the latest games. Blowing me a way quite a bit, to be honest. Really unbelievable the visuals on these titles.
I don't know how you didn't notice this in The Last of Us but with DLSS on Quality, while moving your camera a bit faster, there's a lot blur especially around tree foliage (the leaves become a garbled mess). FSR on Quality is a lot sharper when moving the camera fast in this game but has other issues like you guys noticed. Eventually I just lowered the details a bit and prefered to play without any of them using 1440p ultrawide. Something similar but less noticeable in Cyberpunk as well. From my limited testing, DLSS just seems blurrier when moving fast and FSR isn't as much. In the video you guys seem to just test slow moving with almost no camera moving (like a tech demo), but I might be wrong. Don't want to sound like I don't appreciate the work you guys are putting into this but nobody plays games like that. Of course this might just be me or the limited testing I did but you might want to look into it for a future video maybe.
Great analysis! Maybe it's just because I've never played any of these games to begin with (or any AAA game for that matter), but I'm pretty sure I'm better off with an AMD card since I literally cannot tell the difference. Even when FSR "just looks terrible" the graphics look incredible to me ***edit: the 'shimmering' effects were pretty distracting, but that's basically the only artifacts I could see when they were present***.
I don’t think that’s a reasonable conclusion. If you don’t mind the image here, all that means is that you’d be able to go even lower with internal resolution (maybe at 1080p, or even use ultra performance), but, at some point, you WILL mind the image artifacts. As we’ve seen in the video, the differences get larger the lower the internal resolution, so, if you’re willing to go lower, that means the FSR/DLSS difference will be MORE relevant to you, not less. For people who don’t ever go lower than 4k quality, DLSS and FSR are actually very similar, it’s the people who use the lower res settings that have to deal with a larger difference between DLSS and FSR.
When AMD gets closer or if they ever surpass Nvidia's DLSS visual quality and stability we will have a more accurate price to performance analysis. I always use 1440p Quality when available the FPS gains outweight any bit of ghosting that can result , I would like to see if using DLSS quality at 1440p or DLSS performance at 4K can bypass the 8GB VRAM problem if one wants to play at higher than 1080p native.
Usually textures are tied to output resolution, not render resolution, as a texture that'd blurry at native 4k, will be blurry with dlss a 4k. So there shouldn't be much of a vram decrease from dlss or fsr unfortunately.
I'm on neither team's side, but whenever HU starts talking about how this is "Completely unusable" or "Embarrassing for FSR", I literally cannot notice the difference in the slightest
Is there a visible difference in terms of image quality and performance when you compare FSR or DLSS 1080p Quality vs 1440p Performance vs 4K Ultra performance? Because these presets all are 720p native upscaled to a 1080p.
All reconstruktion technologies are basically made to upscale to 4k. you can technically use it with a 1440P monitor, but unless you are really heavily struggling in terms of FPS, you should only use it (with 1080P) if TAA looks absolutetly trash (which it often does because of ghosting). TAA is a legacy technology for a reason. You should not use it if you are still using a 1080P monitor (it was time to upgrade to 1440P a couple years ago, it's time to upgrade to 4K by now). You can't upscale to 1080P with basically no data which 720P to 1080P is.
@@TheNerd Here in Eastern Europe, i payed 260$ for a 1080p monitor (asus vg249), one year ago, if i was to get a 1440p it would have been 420$, and it would have made no sense with rtx3060 which still costs 500$. I wouldn't run 1440p with anything lower than rtx4070, and that GPU is going to cost around 900$.
@@TheNerd btw Nvidia DLAA is insane on a 1080p monitor, in Last of US, Spiderman, and any other game, it shows that the problem is not with the monitor resolution, it can look amazing, it's about TAA and how many game effects are rendered at half or even quarter resolution compared to native. One ridiculous example is fire in Far Cry 6, that is low resolution, but it looks great in 10 year old games.
Something I've noticed using FSR at 1440p (no DLSS support on my GPU) - if I set my output resolution to 4k using AMD VSR and run ultra performance, the game is scaling from 720p up to 4k, then back down to 1440p. The image produced that way actually looks better and more stable than the same game at 1440p Performance (which is also scaling up from 720p, but going direct to 1440p). As bizarre as it sounds, FSR seems to do a better job from the same source when it's given a higher target resolution to scale to. I'm assuming this is because it has a larger canvas to paint on to, so to speak, so more freedom to retain small details when working out what to keep and what to discard. Needs more investigation, methinks - if this holds true across more titles, then an easy win for game developers using FSR could be to give it a larger canvas and scale that output back down to native. That said, I haven't seen any developers really modify FSR at all - they all use the same fixed render target ratios instead of letting users set custom scaling ratios, for example, even though the latter is supposed to be a supported use case.
@@naturesown4489 Hadn't thought to check that, so just did a run in Cyberpunk in one of the shanty towns with RT turned on. Looks to be a ~5% performance penalty when scaling to 4k instead of 1440p. I'd first noticed the effect when experimenting with 3Dmarks FSR test - with 4k output from same source resolution it just seemed so much sharper, which I hadn't expected.
@@winterscrescendo Using FSR this way, you are essentially being given a low-cost supersampling, which will reduce shimmering on fine details. A few titles like Ys 8 render at a higher internal resolution and then will downsample to the native resolution. It results in an impressively clean image.
@@winterscrescendo That's really cool, I'm going to try it out with Nvidia's DSR or DLDSR (but I don't expect as good performance because DSR has always been worse than VSR).
If you are going to degrade quality by enabling upscaling, and you have to zoom in and really look for a difference. The real conclusion is -> it does not matter at all and will continue to be updated.
I think FSR is amazing for older gaming laptops with Gtx cards, on 15'6 inch screen is a harder to spot any issuse and extra fps boots can really improve expierience.
I'm more interested in performance differences than how it looks. For folks worrying about how their games look to this extent would probably not have any upscaling on to begin with.
After trying both extensively on my RTX 3070, I can honestly say that I think both look terrible and I would rather just drop the settings a touch than deal with weird "blurryness" that both give off.
Great comparison. I really wish nVidia would use the DLSS Preset F (as this is so much better at reducing/removing ghosting) and has slightly superior edge anti-aliasing with only a minor reduction is sharpness compared to the other DLSS Presets. Preset F is only really used for Ultra Performance or DLAA in newer DLSS dll versions. I wish they would make it the standard across all DLSS settings.
You guys are crazy. I don’t know how you continually put out these massive tests, but I’m here to watch them every time. Please don’t work yourselves into the ground though. Take care of your mental health!
I always replace the dlss dll with 2.5.1 and it fixes almost every issue I've ever had with it. Ghosting is almost entirely solved, it's actually incredible. I just wish dlss came with native res built in because most TAA is awful and DLAA is rarely ever used.
Personally, I'd much rather just suck it up and reduce the render resolution than turn on upscaling, because artefacts and stuff that shouldn't be there bother me a lot more than just a generally less sharp image, but when I showed my dad FSR on Amid Evil's ray tracing mode, even I can't ignore that these kind of upscalers are useful. Moreover, he hardly seemed to notice, so I can definitely say that even when FSR isn't as good, the layperson mostly doesn't notice. We can compare resolution numbers and how we prefer to play our games as enthusiasts until the cows come home, but I think it's important to remember that an average person who just got a PC to play some CoD is just going to run a game and turn upscaling on quality if they can't quite get the game to run how they like it and be happy, no matter which upscaler they use. As an aside, I'd be interested in seeing an FSR vs native TAA comparison. I was playing Dead Space and I couldn't stand the gross soupy swimming effect on some of the reflections, so I tried FSR Quality mode and found that it handled those reflections significantly better. I'd be interested in seeing if more games act like that, and if there's an answer to the menace that is mandatory TAA and undersampling.
The YT compression is really making it hard to see a lot of difference, but DLSS with foliage vs FSR really seems like looking through polarized lenses vs regular ones. Impressive amount of testing though. Overall though it doesn't seem like your opinion of FSR has changed much if at all since it launched.
wow awesome all round. DLSS is the ONLY thing keeping me to NVIDIA atm. but the ram issue is making want AMD, im soo torn! question : Do you think DLSS should be a decision maker between NVIDIA and a 6900 or a 7900 series AMD, considering the Ram issues coming over the hill? Cheers guys keep up the good work and testing.
Honestly fsr doesnt look bad at all to me. I do gaming on my 4k mini led tv and i cant tell the difference between native and fsr in most games i play except like dying light 2 but most of that comes down to the games taa implementation I believe
Now redo the FSR with an AMD card and see if it is better. Using both techs on an Nvidia card doesn't paint the whole picture. FSR may perform better on AMD cores.
17:35 There's a fairly simple fix for when the image is too soft, with nvidia at least. Just open the game filters from geforce experience and put the sharpness to something like 10-15% That's usually what i do when using DLSS and getting a too soft result, it does the trick quite well.
@@rafawroblewski8472 I use 1440p and during gaming I can't see the RT unless I am actually looking for it. I can feel the dlls and fsr impact tho. It is even hard to explain what exactly but the image is worse. Some glitches that should not be there. As for TAA it depends on the game, mostly not using it too.
great to see such a chart of what can only be assessed subjectively. this shows that we can definitely measure things like this, even if they become increasingly obfuscated
Why would you want a subjective assessment on a comparison like this? Subjective comparisons don't benefit anyone. And I do believe this comparison is a little subjective.
Have to completely agree with HUB here, I've tried to switch between DLSS and FSR in several games and I never chose FSR over DLSS...it was close in a few instances but NEVER was FSR better than DLSS, which is exactly what this video says
I don't really agree with your conclusion. Yes, DLSS is marginally better, but often you have to nitpick and set them side by side, to see which is better. In a vacuum, where you only have one of them to see, my opinion is that most people will be satisfied with the result, either way. Especially at 4K quality. I've used FSR in a number of games so far and been very happy with the image quality, and been able to reach 60 FPS at 4K, with a 6700XT. I know I'm AMD biased, but i still think it's valid, that if I'm happy with my result, I think a lot of people would agree. My point is that upscaling should not be what affects your buying decision, as you will probably be happy with either, but instead should focus on other features, like performance and cost.
I think it's a lot worse if a game has DLSS but not FSR than vice versa, since FSR is available to every vendor's cards. Using FSR on a RTX card is much preferable to having zero options on a RX card - especially since most RTX cards are powerful enough to run FSR at a high resolution / quality anyway. Of course it's understandable when it happens, since Nvidia sells the vast majority of cards, but then it's just further artificially tilting the playing field to the dominant player's advantage.
4:58 On DlSS you have something resembling TAA filtering (ghosting) Look at the bottom leaves with light as background . I do prefer to loose couple of pixels and some graphycs instead than using upscaler . This is what i learned from this video !
Thanks for the video! I think that DLSS and FSR are mostly relevant for playing at 4K. If you are playing at 1080p - 1440p, just use the native resolution and lower other settings for more performance
1440p DLSS Quality perfectly fine IMO, so far the best DLSS version is 2.5.1 and I use that for every game that I play, this is also an advantage for DLSS because you can easily swap the DLSS file, just a copy paste. You can't do that with FSR.
AMD fanboys will look at this and still wonder why people rather buy NV GPUs. No way I'm touching an AMD card, especially since I play at 1440p. Red fanboys will say they are purists, and that they play at native anyway, but they're deluding themselves. Most modern games require upscaling, especially if you're using a midrange card. And we're not even factoring in ray tracing, I'm talking about just raster games. whether we like it or not, upscaling is here to stay, and only one company does it right.
Again, fail to mention the fact that quality mode on FSR is running at a higher native resolution and therefore the performance and quality comparison is not apples to apples.
Once you find a bargain on some more powerful hardware you can do a better upscale and you wont be able to tell its not 4k from a decent viewing distance give or take certain games
They don’t lean on it. It’s just a way to keep you needing to buy cards sooner. It doesn’t really lower vram usage overall in most games. Imagine a 4070ti with 20gb of vram. It would be fine for years.
What pisses me off is that games today don't look THAT much better than say 2 years ago. Yet, we have to use every trick in the book, down to rendering games at 1080p and upscaling to 4k. Devs are much more lazy now than they used to be imo.
Unfortunately RU-vid compression destroys this comparison. It really needed to be zoomed in to see differences, especially given at least half the audience is watching on their phones.
I hope for a future that both upscaling options are so good that there is no drawback to using them. My guess is when the new GPU gens are coming to the market in around ~2 years. With the current GPU prices DLSS and FSR are a godsend for the longevity for older GPUs from the RTX 2000 and 3000 gens and their AMD counterparts. Also hope that CA implements DLSS and FSR into TW:W3 but that is probably a pipe dream 😂
They already are, those issues are minor, at least the ones on quality mode. You not looking for fine details while shooting enemies or driving in high speed. I gonna be honest, when I actually sit down to play and used those technologies i didn't saw a difference, i was playing on 4k FSR quality, keep that in mind.
Would love to see a test if even modded DLSS is superior to native FSR as well as tests on DLAA native res vs native res. currently playing re4 remake in modded dlaa 4k at it looks way better than everything else in that game
RE4R is pretty much broken when using either FSR or the interlacing mode. A very bad implementation on PC, over on console image scaling is actually okay for whatever reason 😅
No, the only way it could possibly work is if only games that could be modded to the latest version of BOTH were tested with that. But that isn't a lot of games. Nor is modding common, or even allowed in some games, just nothing can be done about "joe homeowner" doing it.
@@Shadowknight890_ PureDark did a few mods to add DLSS to games that doesn't support it natively. Skyrim was his first one & is available on NexusMods.
I disagree that upscaling technology should be a factor when buying a GPU. The image quality is almost guaranteed to be worse regardless of which upscaling is used, so buy your GPU based on other reasons (amount of RAM, overall performance, price). Upscaling should only be used if absolutely necessary.
I must admit that I really don't see much difference especially while I am playing but recognize it is a reality for many people. Thank you for all your detailed work! At least my lack of nuanced vision allows me to buy the best deal card and not worry about all the rest. I should say that I do see much better performance on the lower resolutions with DLSS 2 but wouldn't use upscale on anything less than quality for either technique, so it is a hollow victory for my situation.
2:14: While explaining the scaling factors for the Quality preset for FSR2/DLSS2, which is 67% rez scale, you chose to show in the video Deathloop, the ONLY game that has FSR2 (dunno about DLSS2) Quality preset at 85% rez scale. Just an interesting tidbit.
On the other hand Deathloop is still stuck at FSR 2.0 which, as I commented in another thread, is very old/outdated and not a fair representation of FSR 2 anymore. It was indeed one of the debut games for FSR 2, which despite the version number is "v1.0" software, sharing little with FSR v1.x which was mostly a stopgap solution.
They won't do that tho as they are total AMD fan boys. Not like doubling the FPS isn't a huge factor when choosing a card. I bet you any money they start reviewing with FSR3 when it comes out tho.
I have a 7900XTX. I only use FSR in MW2 because I thought it was worth it. In other titles, like Cyberpunk, I render in native. This video confirmed my suspicions, hopefully AMD catches up in the coming years. And for sure my next GPU purchase will have the upscaling tech in mind. Hopefully that won't be for awhile though. Great video!
I had a Xtx for a bit (untill I found out that mine was gimped with 110deg t junction - damn reference cards!!!) and the lack of DLSS was a little bit of a sore point - however at the moment the brute force will negate that. Quality card no doubt 💪
What I notice from this is how much my eyes/brain combo rejects the artefacts. I really have to look for them to see them. I do wonder if this is because I did a lot of gaming before anisotropic filtering and available anti-aliasing were a thing. Integrated mip-map stuff in gpus was late 90s iirc and usable anti-aliasing was 2008-9? Is peoples ability to filter this stuff out related to how many games they’ve played with these issues just natural to the implementation? I wonder if there’s an age split.
Yesterday I start to use FSR2 with ACC in quality mode, and the amount of FPS more is amazing! I was playing at 4K with epic settings and without FSR2 i was running at 26FPS with FSR2 it went to 39FPS in a VEGA64 I really like it!
@@paulvancyber1979 okay thats what i thought i just couldn’t remember what that other c stood for lol. I just like assetto corsa for the mods as i mainly do touge tracks! Very great game for whatever case although the collision physics can use a little updating.
The upscaling that relies on dedicated hardware is better? What a shock. Again, HUB tests a bunch of games I don't play and won't prioritise when looking for an upgrade, so I'm just looking at raw raster performance and power consumption, as I always do.
Well, HUB doesn't know your game preferences and can't cater only to you. They tested a decent sample and based on their experience selected a few games to show the general trends they see when comparing both technologies. Expecting them to analyze all games that use both technologies or specifically the games you play is just unrealistic and unreasonable.
Hate to ask this, but considering the performance difference between Intel Arc XeSS and other GPUs, is there an image quality difference as well? I don't remember if that was ever covered in the "first look" videos of XeSS. I was also curious if this is the case with FSR as well. Which is to say, does FSR look better with AMD hardware than with other GPUs? I know that's hard to test considering it's not apples-to-apples like using the 4080 in this video...
No, FSR should most probably look the same with AMD or non AMD GPUs. Maybe a slight difference in performance, but that's about it. XeSS on the other hand has two versions. One is for Intel ARC GPUs that utilises the XMX engine, while the other version uses DP4a instructions which can be handled by most modern GPUs. Naturally, the version that relies on Intel's hardware produces a better image (and maybe better perf).
XeSS looks so bad on my Fury X that it's basicly unusable in Warzone 2, on any quality setting. Literally worse then FSR 1, it's hard to describe, blurry and oversharpened at the same time, gives the game a low polygon look.
Great to see see the differences properly demonstrated. The only game I tried FSR on my 6700XT with was Assetto Corsa Competizione and it was barely noticeable for my untrained eye. And I don't think I'm gonna need it for the foreseeable future because I just ordered a 6950XT new for cheap(well at least locally here). Gotta love me those clearance sales!
Yeah, for me--older, bifocals, I don't see that well to begin with--I don't see much difference and kept muttering, "I can't tell the difference." Where I could, it didn't bother me. To me, six on one, half dozen on the other, buy either and I'd be happy with it. Which is good, because I just ordered a 7900 xtx.
The only problem I have with this is you are using a different version of DLSS in every game. All my games now use 2.5.1 which makes a noticeable difference over all the older versions and drops right into any game that uses DLSS 2. In fact in both Witcher 3 and Cyberpunk 2077 DLSS 2.5.1 Quality setting looks better than the native TAAU with virtually none of the ghosting TAAU has plus a nice little performance boost. Being able to easily replace old versions of DLSS in any game that support DLSS 2 is one of the most underrated features of DLSS 2 Only using the version of DLSS 2 available when the game came out is like saying I'm only going to use the graphics driver version that was available when the game released rather than the latest available version ......
My experience with all temporal upscaling have been quite negative, I tried all options available in Spider-man, Witcher 3, COD MW2 and Hogwarts Legacy and they all looked horrible in motion for me, I would rather lower the settings and play at native resolution or just use normal spatial upscaling over dealing with the artifacts that are visible to me when using temporal upscaling. I'm pretty sensitive to those motion artifacts, so I always expected to prefer spatial upscaling over temporal ever since DLSS first came out, but i didn't expect for the difference to still feel this big for me.
That amount of testing must of taken an age but its very interesting to see the comparison especially as I use FSR. I'm very happy with using FSR and in my honest opinion a lot of the time I'm so focused on what I am doing in game I don't really notice anything the only time I do is in certain titles where there is some shimmering or ghosting which can be off putting but that's not very often. It will be interesting to see where FSR 3 goes once that gets implemented.
Since it's basically established that comparing like-with-like settings DLSS always beats out FSR2, especially at lower resolutions, I'd be curious to see how far that advantage takes it. Like, a comparison among multiple games of how FSR2 Performance mode compares with DLSS Ultra Performance mode. Or FSR2 Quality vs DLSS Performance.
Everytime a game has both option, switching between the two, DLSS is always better looking. The sharpening looks cool when you're face first into objects in Cyberpunk, but overall the flickering ruins the experience. Your peripheral vision is more likely to pick up flickering movement over ghosting. So overall, ghosting looks wierd, but it doesn't pull you of of the game as badly as things flickering and shimmering everywhere.
Ghosting is bit more situational (in some games you really won't care much), in CB specifically ghosting related to water and objects looks AWFUL. It's unspeakably ugly. Moving water surface + moving object + upscaler makes almost greasy vomit on screen.
That’s the larger problem for me. If you show me a close up of a ghosting artifact vs a shimmer artifact, they look comparably bad, but when actually playing the game, and both of them taking a very small percentage of your screen, the one that wins is whichever doesn’t draw your attention to it. And in that sense, shimmering and temporal instability is simply the much worse problem to deal with.
Crazy amount of testing... And although I agree with the conclusion, I actually really don't care that much about which tech I use, the flickering of FSR is slightly more distracting that DLSS's ghosting, but ultimately, when I actually get into playing the game at hand I actually forget about the scaling tech. At 1440p Quality (which I prefer) I feel like both techniques offer a native-like experience, with a super sampled look once standing still. :) For me DLSS in itself is not a a sales argument. I will always consider VRAM to be far more important than R&D features.