none of the commenters are watching the clip so I'm just going to summarize and hope someone atleast bothers to read instead. John commented that TSR was the best option for Tekken 8 because it allows 100% resolution scale which means the tech gets to start with native resolution, while DLSS is not given the same option for a full resolution input (DLAA) and instead the best it allows is Quality mode preset which is somewhere around 67% per axis. so while DLSS is being forced to upscale TSR gets to play at native resolution. This does not mean that TSR is categorically better than DLSS as a technology, but that TSR at 100% scale is your best option for maximized image quality specifically in Tekken 8.
Exactly. Unfortunately, as it seems, most people just read provocative titles and react automatically. It's a reflexive behavior... Little thinking involved. 🤷
I went back to DLSS Quality as TSR has A LOT of ghosting/smearing on in game screens, eg on the Coliseum stage. There is more 1-2 pixel thin edge shimmering as well. While actually playing the game, it's very difficult to notice minor differences between the two, but the smearing was just too ugly. IIRC reflections in the floor also looked more smeary with TSR.
TSR is designed by Epic to work particularly well with Nanite, where the upscale is seamless and it had probably to do with its cluster rendering. It's also native, unlike DLSS and FSR that are plugins and they can work only with a limited amount of info from the engine.
Epic should add native support for DLSS, XeSS and FSR2 in Unreal Engine 5, without using external plugins (because plugins have some limitations). Ghostwire Tokyo is UE4 game, and DLSS works better than TSR there. Im not sure if TSR in ue4 is native or bring via plugin, but DLSS is just better there. In ue5, tsr just have tight/native integration with the engine. In Tekken 8 case, we dont have DLAA there, so DLSS quality (67% res scale) vs TSR 100% res scale, then TSR with higher internal resolution is better. But it doesn't mean that TSR is better over DLSS at all.
@@BlackParade01 I know NvRTX. Its Nvidia version of UE5. With some tweaks, improvements etc. Ray Tracing in Nvidia branch runs better and looks better. Im not sure, but probably Wukong will use Nvidia branch. But sadly most devs dont use this version and probably dont know that NvRTX exist. I hope more devs will use it.
You can easily force higher rendering resolution for tekken 8 with "[ConsoleVariables] r.ScreenPercentage=X" command in engine.ini file. This way DLSS Quality can look better than TSR.
Wow , now companies are putting more resources developing more upscalers then better GPUs . Nvidia and amd could have worked harder to give us chiplet GPUs by now ,2x performance increased but they all went the upscalers route . Shit corporations as usually.
no upscaling of any kind at 4k 100% scale is better than TSR 100% scale in tekken 8 john is just wrong on this one look at the arena stage with the screen in the background showing the fighters in real time for testing.
Funny that, i played First Descendant beta, and use the upscale, which the game had three (FSR,XESS, and inhouse upscale), funny for me XESS perform better on me, minimal stutter, clear quality and crisp. While FSR stutters, and blurry. (i'm using 1660 super)
Causes graphics anomalies for me in Tekken 8 especially around the hair and feels like it has noticible input delay. Also Dlaa is easily modded in with Nvidia profile inspector
I’ve played tsr on Fortnite didn’t like it that much bc I noticed certain textures flickering and it’s not the case with dlss, haven’t played it with fsr but most likely will in the future.
TSR is closely integrated with UE5 itself, so it has access to an accurate and larger amount of scene, geometry, and motion vector data compared to FSR 2, DLSS and XeSS. In this regard, DLSS and XeSS has access to AI acceleration to improve image quality to compensate for the lack of extra information that TSR gets, whereas FSR 2 has nothing to compensate this lack of information or accelerate itself with, so that it can work with a larger history frame buffer if possible. Hello Games, closely integrated FSR 2 with their game engine that powers NMS and in that game FSR 2 works really well. Unfortunately, most developers aren't interested in implementing FSR 2/3 properly and just prefer to drag-drop the AMD provided plugin without optimizing the upscaler further according to their games, hence the mediocre results.
The same goes for UE5 games and nanites+lumen. You do not need nanites if it's a small scale low poly game. Abiotic factor threw in lumen which is kinda dumb since the materials and reflections dont even work.
After working in unreal for 14 hours a day for the last 2 months I can tell you... No, TSR is not and will never be better than DLSS. It's better than FSR though, at least visually.
@@marsMayfloweryes you can, use DSR to set the resolution of the game to 1650p and use DLSS to upscale from 1080p, then compare it to TSR in 1080p 100%, or just set TSR to 67% or whatever DLSS quality uses, the options are there, its not that complicated to compare.
Yup that's exactly what I always do on my 1080p monitor. It looks so much sharper and better as compared to DLAA with better performance. It's a win win. P.s. In the game Ark: survival ascended Unreals TSR seems to look better and sharper than DLSS at the same screen percentage. And it also gives much better performance. So using Nvidas tensor cores to upscale using DSR and then using Unreals TSR gives you the best graphics vs performance by far.
@@omarcomming722 I also think they have messed something up, but I'm just stating a fact. TAAU is sharper and more detailed than DLSS in the current game.
That holds true for Dead Space Remake too, TAA on high preset clearly looking better than DLSS quality, I guess it holds differently upon title to title
Probably because TSR is built-in to the engine as opposed DLSS and FSR? XeSS doesnt get enough coverage i think it looks good for PCs that dont support DLSS
I noticed A LOT of visual artifacts with it in Tekken 8 so not sure why John was saying that was the best option to use. Reflections break up with it selected (try the jungle stage), and some edges or textures act weird combined with depth of field (such as Kazuya on the main menu).
One thing to remember is that dlss is a hardware solution, tsr, fsr, and others are software solutions, because of that is almost impossible for those last ones to be as good (idk if xess is hardware as well)
@@DragonOfTheMortalKombat Balanced and Quality presets use a much closer internal resolution than their names would suggest, iirc it's like 840p vs 960p for target 1440p output
I've gotta say that all the talk of upscalers and frame gen I.e. DLSS, FSR, TSR and AI frame gen etc. Aren't these just ways for essentially underpowered GPU's to get to the users desired frame rate!? Just in an artificial way. I wish we could go back to simpler times when everything was native coupled with some AA. If a user wants to improve frame rate, reduce res and settings, nice and simple. Nvidia and AMD are no longer trying to sell us on the power of their GPUs, but on the features, next, they'll be telling us their GPUs have air con and electric windows! All I care about is how fast the engine is, Nvidia / AMD (intel and your XESS, I haven't forgot about you), please keep it simple guys.
What are you talking about? If you want to improve your framerate, you still decrease resolution and settings, it's just that now the image doesn't go immediately to shit when you decrease resolution, it just takes a small hit.
Is there a natural way to gain fps in games? Without software and hardware "tricks" like driver updates or architecture improvements etc. Basically, can I run the pineapple build so I can have my natural way of getting fps? And no, all of these features are about giving choice. Want to sacrifice visual fidelity and gain fps? upscale. Wanna loose on input responsiveness to gain perceived visual smoothness? Frame Gen. It's a choice for every user, from top to bottom tier.
@chacharealsmooth941 That's the problem, i dont really want all this choice. Don't get me wrong, it's my problem and no one elses, if people want choices it's good there are plenty of options. I guess I'm just venting at the sheer amount of 'tricks' that are available to developers that fundamentally, games are now designed in a manner that using these 'tricks' are essential to get decent frame rates. We as gamers are basically forced to choose our poison I.e. I can't play at native, with great image quality and minimal input lag, guess I'll have to give up something on one of those fronts.
Graphics rendering has always been a bunch of "tricks". Any rasterization pipeline in particular is a nested stack of cheats and approximations that has little to do with how light works in reality. I found the best way to look at image reconstruction like TAA is that your brain also does it. Your eye isn't taking a perfectly sharp snapshot every thousandth of a second and copying that buffer into your brain, it's taking many smaller samples over time and having your brain reconstruct them.
@@retroblitz3533 you know, devs were always working with a limited amount of resources, and because of that, always made some compromises. This is just a continuation of that except you can choose not to use it.
if it is hes wrong because bes12000 is correct TSR even at 100scale regardless of what john said has bad artifacting that can be seen on some stages, without any scaling option selected there is no such issue ive tested it, john didnt look at all the stages otherwise he would have noticed this. if you want to test it yourself go to the arena stage that has the screen in the background showing the fighters and watch the crazy movement ghosting that happens its very obvious during gameplay turn TSR 100% scale off and its a non issue.@@iurigrang
it’s actually embarrassing for AMD how bad FSR is. DLSS beat it to market and is still superior, then TSR is also superior with low resolutions, then newcomer intel also beat them with XeSS. what a joke
That tends to be the case. Nvidia always beat AMD to the market with better hardware and software. Buying AMD hardware is a great way to simulate poverty, not much else use for their hardware to be fair.
@@purpasmart_4831 So does TSR, and it looks better than FSR. XeSS also looks better on other GPUs, but runs worse, so TSR is a much better solution and AMD should try getting some answers from epic to improve on FSR, because as it is right now, there's no comparison. It lags too far behind.
see the original tekken or fortnite video for visual showcase. This is DF Clips, and this video is taken from their weekly podcast hence no comparisons.
I've been saying since 2020 that the entire DLSS implementation was piece of trash. The whole concept is lower the internal game resolution to the PS2 era, boost performance, and then try to reconstruct the image back to the monitor main resolution. In other words, they completely distorted the SUPER SAMPLING part of the DLSS acronym and most tech TH-camrs ignored it praising DLSS for ages! Unbelievable!
So you're basically inferring that since 2020 you've not had the first idea what you're talking about? And clearly you know nothing of technical note about the PS2. Best to keep quiet and have a smug look on your face and make people think you know something others don't, rather than open your mouth and make them realise you don't.
@@TheT0nedude What a clown. Maybe you should start use your brain instead of consuming marketing nonsense from tech TH-camrs sponsored by companies that want sell you products. Wanna talk about DLSS? I'm open to show how stupid you have been since 2020 believing that shit.
@@TheT0nedude What a clown. Maybe you should start use your brain instead of consuming marketing nonsense from tech TH-camrs sponsored by companies that want sell you products. Wanna talk about DLSS? I'm open to show how stupid you have been since 2020 believing that sh*t.
@@TheT0nedude What a clown. Maybe you should start use your brain instead of consuming marketing nonsense from tech TH-camrs sponsored by companies that want sell you products. Wanna talk about DLSS? I'm open to show how stupid you have been since 2020 believing that bs.
I don't understand why Richard thinks fighting games are somehow any different than any other game when using an upscaling method. Most game genres are always in motion and including new visual information...
Because big characters that move fast in front of a more static background from a side view create large areas of disoclusion at each frame. He's kind of right, it's a worst case scenario for algorithms that have to guess what was occluded behind the character the frame before at each frame.
@@BlindTrustProject Doesn't seem like a worst case at all. Temporal techniques can guess what the occluded background should be when it is reasonably stable. There are other cases where there is no stable background at all (first and third person games).
@@cube2fox Clearly you are mistaking camera or scenery moving/turning fast where it is easy to extrapole pixels from were they were before on screen thanks to motion vectors, and disocclusion where it is super hard for algo to create pixel from where there was someting completely different the frame before. The worst case scenario here is not the whole screen were it is stable, it is where the big character move around in front of it. For exemple a character that move in front of a backround. If the character takes 10 frames to pass on a given pixel on the screen and suddenly it needs to reconstruct the background instead of the character and all he has is motion vector and colors from the character on previous frame, character that is not there anymore it creates ghosting, trailing and blured contour around the edge of the character. It is a worst case scenario given the size of the character, it makes the disoclusion artefacts very obvious. Thrust me, i'm a veteran game artist and teacher and it is my area of expertise.
Tsr is worse than fsr and dlss. Sometimes devs just sucks or can not optimize quality of dlss and fsr, but they can not make tsr worse since it is special to unreal engine. So we can get games better with tsr but majority of the games, especially if you use 1080p monitor, fsr and dlss is superior. But keep in mind, these guys (df) uses fsr2 with a nvidia gpu while testing it on 4k.
Nah, it's better than fsr2 at least in most games. FSR looks pretty good in stuff like baldur's gate 3 and Avatar, but it looks absolutely dreadful in Alan Wake 2 for example. I have a 6700xt and while I play on a 4k TV, with more demanding games I set my output resolution at 1440p or 1800p. Now, in games I've tried that have TSR and fsr2, tsr usually looks better at balanced mode than fsr2 does at quality with roughly the same framerate. My experience is pretty limited with games that have both (Tekken 8, Ghostwrite Tokyo), but TSR definitely seems to shimmer less while being pretty clear.
@@Torso6131 Thats the problem, tsr loses so much when you upscale to 1080p not 4k. Im a unreal engine dev, I am in love with 100% res tsr, but with 1080p output res, anything below than 80% (which was default ig) it falls apart so much. Fsr2-3 looks just better and more detailed. Especially for texture quality.
Ah, when I was playing on my 1080p TV I tried to avoid using upscaling whenever possible. I couldn't get an acceptable image from FSR at all, and I wasn't impressed with what I saw from dlss either. I would either run games at lower settings or limit my framerate if I had to. I found things for super unstable and ghosty super fast even with fsr2 quality.
@@zap7759 Any resolution has artefacts. That's the nature of a fixed pixel grid because infinite resolution isn't achievable with it. The whole "fake resolution" is such a misnomer that's used by the people that lack knowledge about the ins & outs of what it is they're talking about. Real-time rendered graphics are vastly different from recorded video. As such there might actually be more resolution in a temporal upscaler than there is in a higher output resolution without any temporal sampling being done on its output (because historical frames can have tons of details not being rendered in the current frame thanks to the way real-time graphics are rendered).
none of the commenters are watching the clip so I'm just going to summarize and hope someone atleast bothers to read instead. John commented that TSR was the best option for Tekken 8 because it allows 100% resolution scale which means the tech gets to start with native resolution, while DLSS is not given the same option for a full resolution input (DLAA) and instead the best it allows is Quality mode preset which is somewhere around 67% per axis. so while DLSS is being forced to upscale TSR gets to play at native resolution. This does not mean that TSR is categorically better than DLSS as a technology, but that TSR at 100% scale is your best option for maximized image quality specifically in Tekken 8.
I guess he don't try to do the virtual resolution to match DLSS native then downscale it?
No.
Exactly. Unfortunately, as it seems, most people just read provocative titles and react automatically. It's a reflexive behavior... Little thinking involved. 🤷
I went back to DLSS Quality as TSR has A LOT of ghosting/smearing on in game screens, eg on the Coliseum stage. There is more 1-2 pixel thin edge shimmering as well. While actually playing the game, it's very difficult to notice minor differences between the two, but the smearing was just too ugly. IIRC reflections in the floor also looked more smeary with TSR.
Doesn't TSR at 100% create an overhead instead of increasing framerate? It sounds like it's just TAA
TSR is designed by Epic to work particularly well with Nanite, where the upscale is seamless and it had probably to do with its cluster rendering.
It's also native, unlike DLSS and FSR that are plugins and they can work only with a limited amount of info from the engine.
TSR has definitely impressed. It’s on by default (67% scale) when working in UE5 and it’s hard to notice the difference
Epic should add native support for DLSS, XeSS and FSR2 in Unreal Engine 5, without using external plugins (because plugins have some limitations). Ghostwire Tokyo is UE4 game, and DLSS works better than TSR there. Im not sure if TSR in ue4 is native or bring via plugin, but DLSS is just better there. In ue5, tsr just have tight/native integration with the engine. In Tekken 8 case, we dont have DLAA there, so DLSS quality (67% res scale) vs TSR 100% res scale, then TSR with higher internal resolution is better. But it doesn't mean that TSR is better over DLSS at all.
NvRTX is an Unreal 5 branch with native support
@@BlackParade01 I know NvRTX. Its Nvidia version of UE5. With some tweaks, improvements etc. Ray Tracing in Nvidia branch runs better and looks better. Im not sure, but probably Wukong will use Nvidia branch. But sadly most devs dont use this version and probably dont know that NvRTX exist. I hope more devs will use it.
It's pretty good in RoboCop and Ghostwire too. Especially compared to FSR2
Even xess works better than FSR2.
FSR is garbage compared to any other upscaler.
I wish games were built without the need for scaling, I MUCH prefer the look of native resolution.
Yeah sure, but technology can't handle native res without it costing an arm and a leg.
Nooooo, you are a nVidia hater how dare you to say native looks better than upscaled
@@E-087 loser
honestly native isn't worth the performance cost i prefere to use an upcaler and increase other quality settings
@@Daburademodefinitely! The gain from upscaling can be invested into something that makes a way bigger difference than native res
I play Fortnite and Slender the Arrival Remake with TSR and I keep seeing ghosting behind trees and in tree leaves which I never see with FSR 2.
Nanites are not for low LOD though, quad overdraws are a problen in smaller scale projects vs manual LOD which takes minutes.
Look up Threat Interactive and the folly of UE5 for basic games...
Better off using ue4 cuz its optimized.
Doesn't DLSS perform a lot better than TSR? So why would you even pick TSR when you can get much better graphics/framerate with DLSS?
You can easily force higher rendering resolution for tekken 8 with "[ConsoleVariables] r.ScreenPercentage=X" command in engine.ini file. This way DLSS Quality can look better than TSR.
It's not even close to DLSS even at 4k, I'm surpirsed how positive they are on it because it looked like shit to me even at quality tier.
Could you list all of the upscale... 🎉❤❤
why are we internally rendering games below 1080p? Thats the real question. Deffered rendering has gone too far.
Wow , now companies are putting more resources developing more upscalers then better GPUs .
Nvidia and amd could have worked harder to give us chiplet GPUs by now ,2x performance increased but they all went the upscalers route .
Shit corporations as usually.
no upscaling of any kind at 4k 100% scale is better than TSR 100% scale in tekken 8 john is just wrong on this one look at the arena stage with the screen in the background showing the fighters in real time for testing.
Funny that, i played First Descendant beta, and use the upscale, which the game had three (FSR,XESS, and inhouse upscale), funny for me XESS perform better on me, minimal stutter, clear quality and crisp. While FSR stutters, and blurry. (i'm using 1660 super)
You can easily mod in dlaa to tekken or any dlss game
Yeah this is what I do. It literally takes a few seconds.
@@via_negativa6183 too bad it's a bit harder to remove taa
Doesn't that get you banned?
@@mat_max lmao no
@@GFClockedhow? Are you talking about dlaa or dldsr?
Causes graphics anomalies for me in Tekken 8 especially around the hair and feels like it has noticible input delay. Also Dlaa is easily modded in with Nvidia profile inspector
I’ve played tsr on Fortnite didn’t like it that much bc I noticed certain textures flickering and it’s not the case with dlss, haven’t played it with fsr but most likely will in the future.
TSR is closely integrated with UE5 itself, so it has access to an accurate and larger amount of scene, geometry, and motion vector data compared to FSR 2, DLSS and XeSS. In this regard, DLSS and XeSS has access to AI acceleration to improve image quality to compensate for the lack of extra information that TSR gets, whereas FSR 2 has nothing to compensate this lack of information or accelerate itself with, so that it can work with a larger history frame buffer if possible.
Hello Games, closely integrated FSR 2 with their game engine that powers NMS and in that game FSR 2 works really well. Unfortunately, most developers aren't interested in implementing FSR 2/3 properly and just prefer to drag-drop the AMD provided plugin without optimizing the upscaler further according to their games, hence the mediocre results.
DLSS is an Unreal Engine plugin and can access the same buffers as TSR.
The same goes for UE5 games and nanites+lumen.
You do not need nanites if it's a small scale low poly game. Abiotic factor threw in lumen which is kinda dumb since the materials and reflections dont even work.
After working in unreal for 14 hours a day for the last 2 months I can tell you... No, TSR is not and will never be better than DLSS. It's better than FSR though, at least visually.
What game are you working on?
how about watching the clip first?
Cool, do you work on rendering pipeline?
Kind of hope to hear more details from you.
You can basically test by setting DLSS to higher target resolution so that the internal res matches, then downsample
but you can't set DLSS to native res, only lower than native, so no way to compare TSR at 100% res.
@@marsMayfloweryes you can, use DSR to set the resolution of the game to 1650p and use DLSS to upscale from 1080p, then compare it to TSR in 1080p 100%, or just set TSR to 67% or whatever DLSS quality uses, the options are there, its not that complicated to compare.
@@marsMayflower You can use DLDSR or DSR (the thing which, for example, lets you play in 4K on 1080p monitor) with DLSS.
Yup that's exactly what I always do on my 1080p monitor. It looks so much sharper and better as compared to DLAA with better performance. It's a win win.
P.s. In the game Ark: survival ascended Unreals TSR seems to look better and sharper than DLSS at the same screen percentage. And it also gives much better performance. So using Nvidas tensor cores to upscale using DSR and then using Unreals TSR gives you the best graphics vs performance by far.
@@nicolass4296 ah, yea, forgot about the DRS stuff. kind of a wonky way to get there, but hey, it works
XeSS is very heavy on ancient hardware compared to other methods.
TAA upscaling @ quality preset looks better than DLSS @ quality in Skull & Bones.
No chance unless they're using a fcked up old dlss dll with forced sharpening/smoothing, or they completely messed up the implementation somehow
@@omarcomming722 I also think they have messed something up, but I'm just stating a fact.
TAAU is sharper and more detailed than DLSS in the current game.
No one cares about Dull and Bones
@@Dr.WhetFarts sick burn bro
That holds true for Dead Space Remake too, TAA on high preset clearly looking better than DLSS quality, I guess it holds differently upon title to title
Probably because TSR is built-in to the engine as opposed DLSS and FSR? XeSS doesnt get enough coverage i think it looks good for PCs that dont support DLSS
Problem is that XeSS usually has the worst performance out of the upscalers.
Exactly. I believe TSR is very much intertwined in the way unreal builds up its frames. It leverages the way they create geometry/meshes for example.
No, it's because tsr is at native resolution, dlss isn't.
and its not open source...so very hard to integrate@@Kait0s
@@iurigrang TSR 100% res scale acts like DLAA, u can change the res scale to whatever u want in most UE games
I noticed A LOT of visual artifacts with it in Tekken 8 so not sure why John was saying that was the best option to use. Reflections break up with it selected (try the jungle stage), and some edges or textures act weird combined with depth of field (such as Kazuya on the main menu).
prefer DLAA over all
Can confirm TSR is better than FSR 3.0 and XeSS 1.3 appearance and performance wise in Hellblade 2
DF are the last people still playing Immortals of Aveum.
One thing to remember is that dlss is a hardware solution, tsr, fsr, and others are software solutions, because of that is almost impossible for those last ones to be as good (idk if xess is hardware as well)
Xess is hardware as well ( on intel GPUs) . Xess balanced has comparable image quality to fsr quality mode.
@@DragonOfTheMortalKombat Balanced and Quality presets use a much closer internal resolution than their names would suggest, iirc it's like 840p vs 960p for target 1440p output
@@DragonOfTheMortalKombat isn't that the case on the dp4a path? I was under the impression actual Intel gpu xess was even better
There are two different versions of XESS, one hardware, one software. Naturally, the former is much better and exclusive to Intel Arc.
Screw Epic, UE sucks - they're trying their best to get big titles to use it.
And they are..
be-he-moth?
Yes for image quality but not performance....
It's actually (almost) free when talking about performance, whereas DLSS is not. This was mentioned in the video.
@@DeltaNovumRunning the game at 100% resolution certainly isn’t free, even if the TSR algorithm itself is efficient.
I've gotta say that all the talk of upscalers and frame gen I.e. DLSS, FSR, TSR and AI frame gen etc. Aren't these just ways for essentially underpowered GPU's to get to the users desired frame rate!? Just in an artificial way. I wish we could go back to simpler times when everything was native coupled with some AA. If a user wants to improve frame rate, reduce res and settings, nice and simple. Nvidia and AMD are no longer trying to sell us on the power of their GPUs, but on the features, next, they'll be telling us their GPUs have air con and electric windows! All I care about is how fast the engine is, Nvidia / AMD (intel and your XESS, I haven't forgot about you), please keep it simple guys.
What are you talking about? If you want to improve your framerate, you still decrease resolution and settings, it's just that now the image doesn't go immediately to shit when you decrease resolution, it just takes a small hit.
Is there a natural way to gain fps in games? Without software and hardware "tricks" like driver updates or architecture improvements etc. Basically, can I run the pineapple build so I can have my natural way of getting fps?
And no, all of these features are about giving choice. Want to sacrifice visual fidelity and gain fps? upscale. Wanna loose on input responsiveness to gain perceived visual smoothness? Frame Gen.
It's a choice for every user, from top to bottom tier.
@chacharealsmooth941 That's the problem, i dont really want all this choice. Don't get me wrong, it's my problem and no one elses, if people want choices it's good there are plenty of options. I guess I'm just venting at the sheer amount of 'tricks' that are available to developers that fundamentally, games are now designed in a manner that using these 'tricks' are essential to get decent frame rates. We as gamers are basically forced to choose our poison I.e. I can't play at native, with great image quality and minimal input lag, guess I'll have to give up something on one of those fronts.
Graphics rendering has always been a bunch of "tricks". Any rasterization pipeline in particular is a nested stack of cheats and approximations that has little to do with how light works in reality.
I found the best way to look at image reconstruction like TAA is that your brain also does it. Your eye isn't taking a perfectly sharp snapshot every thousandth of a second and copying that buffer into your brain, it's taking many smaller samples over time and having your brain reconstruct them.
@@retroblitz3533 you know, devs were always working with a limited amount of resources, and because of that, always made some compromises. This is just a continuation of that except you can choose not to use it.
Tsr is way better than Fsr but not dlss.
I test 50% scale on 4K, it is definitely worse than DLSS.
NO upscaling at all with resolution at 100 and everything on ultra is the best way fro Tekken 8.
That's... What he is talking about? Watch the video.
No the best way is to use DLAA in 4K
if it is hes wrong because bes12000 is correct TSR even at 100scale regardless of what john said has bad artifacting that can be seen on some stages, without any scaling option selected there is no such issue ive tested it, john didnt look at all the stages otherwise he would have noticed this. if you want to test it yourself go to the arena stage that has the screen in the background showing the fighters and watch the crazy movement ghosting that happens its very obvious during gameplay turn TSR 100% scale off and its a non issue.@@iurigrang
"Maybe DLAA would have an edge" - nice one, John!
In other words, DLSS > TSR > FSR2 when upscalling from same resolution
XeSS beats FSR in numerous games too.
it’s actually embarrassing for AMD how bad FSR is. DLSS beat it to market and is still superior, then TSR is also superior with low resolutions, then newcomer intel also beat them with XeSS. what a joke
That tends to be the case. Nvidia always beat AMD to the market with better hardware and software. Buying AMD hardware is a great way to simulate poverty, not much else use for their hardware to be fair.
@@fcukugimmeausernameAMD GPU are good for open-source. The Linux community loves them.
FSR runs on any gpu though.
@@purpasmart_4831 So does TSR, and it looks better than FSR. XeSS also looks better on other GPUs, but runs worse, so TSR is a much better solution and AMD should try getting some answers from epic to improve on FSR, because as it is right now, there's no comparison. It lags too far behind.
@@viniqf TSR is only UE5 though and FSR is actually open source.
Lots of acronyms here guys 😐
The expectation is you click on a technical video knowing what they are or are curious to learn more. Stop whining.
How many years away are we from such a day where all this tech just works equally as the human eye is concerned and such debates fade away.
No.
Summery
TSR isn't better than DLSS. Just another confused person asking questions
Show us difference instead of bla-bla-bla.
see the original tekken or fortnite video for visual showcase. This is DF Clips, and this video is taken from their weekly podcast hence no comparisons.
I've been saying since 2020 that the entire DLSS implementation was piece of trash. The whole concept is lower the internal game resolution to the PS2 era, boost performance, and then try to reconstruct the image back to the monitor main resolution.
In other words, they completely distorted the SUPER SAMPLING part of the DLSS acronym and most tech TH-camrs ignored it praising DLSS for ages! Unbelievable!
So you're basically inferring that since 2020 you've not had the first idea what you're talking about? And clearly you know nothing of technical note about the PS2. Best to keep quiet and have a smug look on your face and make people think you know something others don't, rather than open your mouth and make them realise you don't.
@@TheT0nedudeI would have not phrased it better.
@@TheT0nedude What a clown. Maybe you should start use your brain instead of consuming marketing nonsense from tech TH-camrs sponsored by companies that want sell you products.
Wanna talk about DLSS? I'm open to show how stupid you have been since 2020 believing that shit.
@@TheT0nedude What a clown. Maybe you should start use your brain instead of consuming marketing nonsense from tech TH-camrs sponsored by companies that want sell you products.
Wanna talk about DLSS? I'm open to show how stupid you have been since 2020 believing that sh*t.
@@TheT0nedude What a clown. Maybe you should start use your brain instead of consuming marketing nonsense from tech TH-camrs sponsored by companies that want sell you products.
Wanna talk about DLSS? I'm open to show how stupid you have been since 2020 believing that bs.
I don't understand why Richard thinks fighting games are somehow any different than any other game when using an upscaling method. Most game genres are always in motion and including new visual information...
Because big characters that move fast in front of a more static background from a side view create large areas of disoclusion at each frame. He's kind of right, it's a worst case scenario for algorithms that have to guess what was occluded behind the character the frame before at each frame.
@@BlindTrustProject Doesn't seem like a worst case at all. Temporal techniques can guess what the occluded background should be when it is reasonably stable. There are other cases where there is no stable background at all (first and third person games).
@@cube2fox Clearly you are mistaking camera or scenery moving/turning fast where it is easy to extrapole pixels from were they were before on screen thanks to motion vectors, and disocclusion where it is super hard for algo to create pixel from where there was someting completely different the frame before.
The worst case scenario here is not the whole screen were it is stable, it is where the big character move around in front of it.
For exemple a character that move in front of a backround. If the character takes 10 frames to pass on a given pixel on the screen and suddenly it needs to reconstruct the background instead of the character and all he has is motion vector and colors from the character on previous frame, character that is not there anymore it creates ghosting, trailing and blured contour around the edge of the character. It is a worst case scenario given the size of the character, it makes the disoclusion artefacts very obvious.
Thrust me, i'm a veteran game artist and teacher and it is my area of expertise.
Interesting
Interesting
off-topic, but the pronunciation of the word comparable as "kumpearuhbul" instead of "kahmperruhbul" is super annoying.
Tsr is worse than fsr and dlss. Sometimes devs just sucks or can not optimize quality of dlss and fsr, but they can not make tsr worse since it is special to unreal engine. So we can get games better with tsr but majority of the games, especially if you use 1080p monitor, fsr and dlss is superior. But keep in mind, these guys (df) uses fsr2 with a nvidia gpu while testing it on 4k.
Nah, it's better than fsr2 at least in most games. FSR looks pretty good in stuff like baldur's gate 3 and Avatar, but it looks absolutely dreadful in Alan Wake 2 for example. I have a 6700xt and while I play on a 4k TV, with more demanding games I set my output resolution at 1440p or 1800p.
Now, in games I've tried that have TSR and fsr2, tsr usually looks better at balanced mode than fsr2 does at quality with roughly the same framerate. My experience is pretty limited with games that have both (Tekken 8, Ghostwrite Tokyo), but TSR definitely seems to shimmer less while being pretty clear.
@@Torso6131you said it, you olay in 4k. On 1080 tsr is extra ghosty and shimmery
@@Torso6131 Thats the problem, tsr loses so much when you upscale to 1080p not 4k. Im a unreal engine dev, I am in love with 100% res tsr, but with 1080p output res, anything below than 80% (which was default ig) it falls apart so much. Fsr2-3 looks just better and more detailed. Especially for texture quality.
Ah, when I was playing on my 1080p TV I tried to avoid using upscaling whenever possible. I couldn't get an acceptable image from FSR at all, and I wasn't impressed with what I saw from dlss either. I would either run games at lower settings or limit my framerate if I had to. I found things for super unstable and ghosty super fast even with fsr2 quality.
wow, people really believe nvidias bull, you all got sold on pseudo hardware driven applicational software LOL
"pseudo hardware driven applicational software", yet, not a single software competidor has managed to match DLSS
@@maldosogamerFake resolutions have artifacts. Probably better to view it as a crutch than "Oh wow, 4K gaming".
@@zap7759 Every AA implementation has artifacts.
That argument is so stupid. You can literally categorize rasterization as "pseudo hardware driven applicational software".
@@zap7759 Any resolution has artefacts. That's the nature of a fixed pixel grid because infinite resolution isn't achievable with it. The whole "fake resolution" is such a misnomer that's used by the people that lack knowledge about the ins & outs of what it is they're talking about. Real-time rendered graphics are vastly different from recorded video. As such there might actually be more resolution in a temporal upscaler than there is in a higher output resolution without any temporal sampling being done on its output (because historical frames can have tons of details not being rendered in the current frame thanks to the way real-time graphics are rendered).