• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD's FSR3 possibly next month ?

Well I'm doubling down on the findings I reported yesterday (post #884 just above) and I'm still calling total Emperor's New Clothes on this mod.

To make sure I'm not going mad and that there's nothing wrong with my screen (4k OLED HDR, 120Hz VRR Gsync compatible) I fired up Dishonoured today which for some reason has always been my go to game for testing refresh rates and smoothness feel. I know this game doesn't have frame gen or anything (it's old!) but I just wanted to double check the look and feel of 120Hz/120fps vs 60Hz/60fps gaming......and it's exactly as it's always been and how it should be! 120Hz @ 120fps has vastly reduced motion ghosting vs 60hz @ 60fps and obviously looks and feels miles better.

In all my testing this week with this mod as described, even though my screen is at 120Hz and the frames are at 116 when game settings are adjusted accordingly (as Reflex is on by default with this mod it won't hit 120) it looks and feels no different Vs running without the mod.
(And I've tested with games running at max possible fps and matching refresh rate so VRR doesn't have to do its thing and with more punishing graphics settings so VRR does do its thing. This is more for frametime findings though rather than motion ghosting, and some decent youtubers have found that if you can't match your fps to your screen Hz then framepacing is still off with FSR3 frame gen for many games.)

The best example I find is Cyberpunk as this has the least issues and supposedly works the best - The amount of motion ghosting is identical whether running at ~70fps without the mod Vs running at 116fps with the mod. It does nothing apart from show a higher frame rate in your osd.
 
Last edited:
TBF, your post isn't exactly clear at what you're trying to get at but streamline itself is not locked down to RTX gpus, in fact, it's not even anything relevant to us end users/gamers, it's purely for developers.


And what g67575 has stated isn't correct, whole point of streamline is to faciliate the implementation of all rtx tech and other branded features in one go as opposed to devs having to do it seperately so blame amd for not wanting to join OSS.



Both are just as bad as each other, neither company is good and it's foolish to think amd are the good guys, they aren't, you buy what fits your needs and budget at the time, don't let morals get in the way as neither company has your best interests at heart.

The reason I say Streamline is locked is because the Nukem9 (the author of the mod) on discord himself said that is the case. The sl.interposer.dll and perhaps some other dll's apparently check your GPU ID before the DLSS or FG options in a game can be selected. Not all of Streamline is open source hence why modders cant make the mod useable for AMD cards.
 
Last edited:
Well I'm doubling down on the findings I reported yesterday (post #884 just above) and I'm still calling total Emperor's New Clothes on this mod.

To make sure I'm not going mad and that there's nothing wrong with my screen (4k OLED HDR, 120Hz VRR Gsync compatible) I fired up Dishonoured today which for some reason has always been my go to game for testing refresh rates and smoothness feel. I know this game doesn't have frame gen or anything (it's old!) but I just wanted to double check the look and feel of 120Hz vs 60Hz gaming......and it's exactly as it's always been and how it should be! 120Hz has vastly reduced motion ghosting vs 60hz and obviously looks and feels miles better.

In all my testing this week with this mod as described, even though my screen is at 120Hz and the frames are at 116 when game settings are adjusted accordingly (as Reflex on by default it won't hit 120) it looks and feels no different Vs running without the mod.
The best example is Cyberpunk as this has the least issues and supposedly works the best - The amount of motion ghosting is identical whether running at ~70fps without the mod Vs running at 116fps with the mod. It does nothing.

Honestly I would say there is something potentially wrong with your setup or/and a setting is conflicting somewhere causing there to be no difference at all as I'm extremely sensitive to stutter, motion fluidity etc. and whilst it isn't as good as a native high fps, it most definetly is improving the motion fluidity and in some cases can improve the clarity of motion too (hogwarts, ark and starfield being the ones to show a notable improvement in motion clarity). I swap out dlss for the 3.5.1 version and sometimes also switch out the preset for a better one such as C, which usually has less ghosting than the D preset so perhaps that could be a contributing factor?

The reason I say Steamline is locked is because the Nukem9 (the author of the mod) on discord himself said that is the case. The sl.interposer.dll and perhaps some other dll's apparently check your GPU ID before the DLSS or FG options in a game can be selected. Not all of Streamline is open source hence why modders cant make the mod useable for AMD cards.

In what specific way is it locked though? As in RTX features being locked based on it detecting rtx gpus? Isn't this what we would expect though? Otherwise people would be able to enable rtx features on non rtx gpus, which we know nvidia don't want.

Perhaps the modders can't make use of streamline for amd gpus because amd didn't onboard/sign up to use it? Thus it can't be used to enable what are essentially RTX specific features?
 
@Nexus18 I'm pretty sure there's nothing wrong with my setup, and I too am ultra sensitive to stutter and motion fluidity, too sensitive really as I waste a lot of time tinkering instead of actually gaming!

I'll try Starfield again to double check for motion clairty/ ghosting differences with the mod on vs off, but this is one game that definitely has worse frame pacing with the mod on vs off, so I would never use it anyway in its current state.

EDIT - just tried Starfield again. Because I was getting worse frame pacing I didn't give it much of a chance when testing, but on looking again - purely from a motion ghosting/ motion clarity point of view - I think the mod is working for this game. ~70fps with the mod off vs ~110fps with it on, and I can see the slight increase in motion clarity/ less motion ghosting being evident, that you would expect and hope for running at higher fps. This is what I'm not seeing in Cyberpunk, at all, which is why it's bothering me.

I also had a custom reshade running on Starfield so I've taken this out and I think this might have helped a touch with the framepacing, so there may be some slight conflict there. I have a few mods running with Cyberpunk, though hardly any, but I fail to see how these could affect the look and feel of motion clarity???

Do you use DLSSTweaks then, sounds like you do? I've been meaning to look into it and give it a whirl....
 
Last edited:
@Nexus18 I'm pretty sure there's nothing wrong with my setup, and I too am ultra sensitive to stutter and motion fluidity, too sensitive really as I waste a lot of time tinkering instead of actually gaming!

I'll try Starfield again to double check for motion clairty/ ghosting differences with the mod on vs off, but this is one game that definitely has worse frame pacing with the mod on vs off, so I would never use it anyway in its current state.

EDIT - just tried Starfield again. Because I was getting worse frame pacing I didn't give it much of a chance when testing, but on looking again - purely from a motion ghosting/ motion clarity point of view - I think the mod is working for this game. ~70fps with the mod off vs ~110fps with it on, and I can see the slight increase in motion clarity/ less motion ghosting being evident, that you would expect and hope for running at higher fps. This is what I'm not seeing in Cyberpunk, at all, which is why it's bothering me.

I also had a custom reshade running on Starfield so I've taken this out and I think this might have helped a touch with the framepacing, so there may be some slight conflict there. I have a few mods running with Cyberpunk, though hardly any, but I fail to see how these could affect the look and feel of motion clarity???

Do you use DLSSTweaks then, sounds like you do? I've been meaning to look into it and give it a whirl....

It's much better frame pacing/motion for me here in starfield.


I only use DLSS tweaks if I notice ghosting so not too often really, I'm not using it for any of the current games where fsr 3 is modded in though (have switched out all the dlss using dlss swapper for version 3.5.10 though). Sadly youtube videos are pretty pointless especially since locked to 60 fps so not getting to see the real benefit of higher fps i.e. motion fluidity but you can somewhat at least see it is less stuttery at times, also shows in the frame time graph, this is where Daniel Owens videos are better as he is recording the screen directly.





I never use reshades or any kind of post processing effect due to the sharpening and so on (even if a minimal amount it), it can enhance visual artifacts that will be in the base/native frame especially when using upscaling tech.

Recall of people saying that ray reconstruction with path tracing can have some ghosting when used with frame gen. in cp 2077. I do agree that the motion isn't as clean in CP 2077 as other titles but I just put this down to 30xx being slower/weaker for PT workload and with a higher FPS, it's more obvious since the stutter/low fps is more prevalent than poor motion clarity, might be worth trying without ray reconstruction and/or normal RT over PT.
 
Last edited:
Honestly I would say there is something potentially wrong with your setup or/and a setting is conflicting somewhere causing there to be no difference at all as I'm extremely sensitive to stutter, motion fluidity etc. and whilst it isn't as good as a native high fps, it most definetly is improving the motion fluidity and in some cases can improve the clarity of motion too (hogwarts, ark and starfield being the ones to show a notable improvement in motion clarity). I swap out dlss for the 3.5.1 version and sometimes also switch out the preset for a better one such as C, which usually has less ghosting than the D preset so perhaps that could be a contributing factor?



In what specific way is it locked though? As in RTX features being locked based on it detecting rtx gpus? Isn't this what we would expect though? Otherwise people would be able to enable rtx features on non rtx gpus, which we know nvidia don't want.

Perhaps the modders can't make use of streamline for amd gpus because amd didn't onboard/sign up to use it? Thus it can't be used to enable what are essentially RTX specific features?

If modders cannot modify the 'opensource' streamline to bypass any checks to allow AMD to select DLSS or FG then it means Streamline is not fully opensource. Being able to select an option in the game does not mean it will run of course. Modders have been able to get FSR3 running with any card that can enable the FG option in the games so there is no reason why AMD could not run that if streamline didn't have hard locks in place.
 
What has that got to do with what I posted? Do you work for Nvidia and feel the need to defend them? If AMD is locking down FSR3 then that is their issue. Nvidia is the pioneer of the proprietary disease since they took over Ageia (Physx) so not sure why AMD is the evil corporation in your eyes.
In any case, Which is worse...Nvidia allowing all upscalers with frame gen on their 4000 cards or AMD locking the upscaler to FSR2 but providing frame gen for all?
No, he just has Jensen pictures on his wall and a big Nvidia tattoo on his back :D
 
Last edited:
It's much better frame pacing/motion for me here in starfield.


I only use DLSS tweaks if I notice ghosting so not too often really, I'm not using it for any of the current games where fsr 3 is modded in though (have switched out all the dlss using dlss swapper for version 3.5.10 though). Sadly youtube videos are pretty pointless especially since locked to 60 fps so not getting to see the real benefit of higher fps i.e. motion fluidity but you can somewhat at least see it is less stuttery at times, also shows in the frame time graph, this is where Daniel Owens videos are better as he is recording the screen directly.

I never use reshades or any kind of post processing effect due to the sharpening and so on (even if a minimal amount it), it can enhance visual artifacts that will be in the base/native frame especially when using upscaling tech.

Recall of people saying that ray reconstruction with path tracing can have some ghosting when used with frame gen. in cp 2077. I do agree that the motion isn't as clean in CP 2077 as other titles but I just put this down to 30xx being slower/weaker for PT workload and with a higher FPS, it's more obvious since the stutter/low fps is more prevalent than poor motion clarity, might be worth trying without ray reconstruction and/or normal RT over PT.

With the mod DLSS and XeSS do to have ghosting, especially on the car tail lights. However, when I selected FSR2 there was almost none (apart from behind the car) which indicates that AMD has indeed optimized FSR3 for their own upscaler. It is obvious they will not optimize for other upscalers. FSR2 does have a lot more shimmer than the native implementation though.
 
Last edited:
If modders cannot modify the 'opensource' streamline to bypass any checks to allow AMD to select DLSS or FG then it means Streamline is not fully opensource. Being able to select an option in the game does not mean it will run of course. Modders have been able to get FSR3 running with any card that can enable the FG option in the games so there is no reason why AMD could not run that if streamline didn't have hard locks in place.
Streamline itself is open source, Nvidia have it advertised as such, if it wasn't, they would be in deep **** for this:


Dlss and other rtx tech is not open source which is why amd and intel can't use it as it is closed source and locked to rtx GPUs only, so streamline is open source but the rtx tech in it is not oss.

With the mod DLSS and XeSS do to have ghosting, especially on the car tail lights. However, when I selected FSR2 there was almost none which indicates that AMD has indeed optimized FSR3 for their own upscaler. It is obvious they will not optimize for other upscalers.

I'll check out fsr 2 later with the mod and confirm how it looks for me as I have noticed ghosting with path tracing, ray reconstruction and dlss and the frame gen mod but without the frame gen, there is little to no noticeable ghosting.
 
I'll check out fsr 2 later with the mod and confirm how it looks for me as I have noticed ghosting with path tracing, ray reconstruction and dlss and the frame gen mod but without the frame gen, there is little to no noticeable ghosting.

Did a bit more testing and the easiest place to see the ghosting is to move the car forward and backwards near some crossing overhead power lines. Both DLSS and XeSS show some ghosting (or frame gen artifacts) on the power lines but FSR 2 doesn't. Overall I'd still play with DLSS since FSR2 has the usuall shimmering issue and more than the native implementation.On my 3080FE I get around 80fps with PT and DLLS Quality + FSR3 FG. It definitely feels a lot smoother than without FG.
 
Last edited:
It's much better frame pacing/motion for me here in starfield.


I only use DLSS tweaks if I notice ghosting so not too often really, I'm not using it for any of the current games where fsr 3 is modded in though (have switched out all the dlss using dlss swapper for version 3.5.10 though). Sadly youtube videos are pretty pointless especially since locked to 60 fps so not getting to see the real benefit of higher fps i.e. motion fluidity but you can somewhat at least see it is less stuttery at times, also shows in the frame time graph, this is where Daniel Owens videos are better as he is recording the screen directly.



I never use reshades or any kind of post processing effect due to the sharpening and so on (even if a minimal amount it), it can enhance visual artifacts that will be in the base/native frame especially when using upscaling tech.

Recall of people saying that ray reconstruction with path tracing can have some ghosting when used with frame gen. in cp 2077. I do agree that the motion isn't as clean in CP 2077 as other titles but I just put this down to 30xx being slower/weaker for PT workload and with a higher FPS, it's more obvious since the stutter/low fps is more prevalent than poor motion clarity, might be worth trying without ray reconstruction and/or normal RT over PT.

Ok, done some further testing in Starfield and have made sure all my custom reshade files are cleaned out. I can happily report now that this specific game really seems like a great example of the mod working as it should/ as I would have hoped. So I'm semi- back on the DLLS to FSR3 frame gen/ mod hype train. What a doofus :D There's clearly an issue with Cyberpunk for me though so I definitely wasn't imagining it, I need to test it more thoroughly.

Also you really shouldn't discount reshade fella, especially as you sound like a fellow tweaker. I never download pre-sets, I only make my own reshades from scratch. 99% of presets out there are ******* garbage made by idiots who have no idea about image quality. Even if you do download pre-sets though you can easily turn off any sharpening that might have been in there. It's a superb tool that you can really fine tune any game with.
 
Last edited:
Did a bit more testing and the easiest place to see the ghosting is to move the car forward and backwards near some crossing overhead power lines. Both DLSS and XeSS show some ghosting (or frame gen artifacts) on the power lines but FSR 2 doesn't. Overall I'd still play with DLSS since FSR2 has the usuall shimmering issue and more than the native implementation.On my 3080FE I get around 80fps with PT and DLLS Quality + FSR3 FG. It definitely feels a lot smoother than without FG.

Tried it myself there and FSR absolutely does have ghosting, in some parts, noticeably worse than dlss and xess:


Ray reconstruction adds to the ghosting/artifacts somewhat. On the car backend, that's the only time in the game that I really notice the artifacts with frame gen on, not even so much the traditional ghosting but just pure artifacts.

Maybe it is better with quality modes but then that's not usable with path tracing at 3440x1440 on a 3080 (well it's not as enjoyable/playable imo). Not sure how you're getting that fps with dlss quality and FG unless you're gaming at 1080P? 2560x1440 is a bit less demanding than 21.9 3440x1440 but still seems awfully high.

Ok, done some further testing in Starfield and have made sure all my custom reshade files are cleaned out. I can happily report now that this specific game really seems like a great example of the mod working as it should/ as I would have hoped. So I'm semi- back on the DLLS to FSR3 frame gen/ mod hype train. What a doofus :D There's clearly an issue with Cyberpunk for me though so I definitely wasn't imagining it, I need to test it more thoroughly.

Also you really shouldn't discount reshade fella, especially as you sound like a fellow tweaker. I never download pre-sets, I only make my own reshades from scratch. 99% of presets out there are ******* garbage made by idiots who have no idea about image quality. Even if you do download pre-sets though you can easily turn off any sharpening that might have been in there. It's a superb tool that you can really fine tune any game with.

I think ever since getting OLED, first LG oled and now my main gaming display, the aw qd-oled, I haven't really felt the need for reshades especially if HDR is supported as everything looks so good as it is :D Used to use them way back in the day on my lcd monitors, especially for games like dishonored which lacked that pop/colour.
 
Tried it myself there and FSR absolutely does have ghosting, in some parts, noticeably worse than dlss and xess:


Ray reconstruction adds to the ghosting/artifacts somewhat. On the car backend, that's the only time in the game that I really notice the artifacts with frame gen on, not even so much the traditional ghosting but just pure artifacts.

Maybe it is better with quality modes but then that's not usable with path tracing at 3440x1440 on a 3080 (well it's not as enjoyable/playable imo). Not sure how you're getting that fps with dlss quality and FG unless you're gaming at 1080P? 2560x1440 is a bit less demanding than 21.9 3440x1440 but still seems awfully high.



I think ever since getting OLED, first LG oled and now my main gaming display, the aw qd-oled, I haven't really felt the need for reshades especially if HDR is supported as everything looks so good as it is :D Used to use them way back in the day on my lcd monitors, especially for games like dishonored which lacked that pop/colour.

I see you picked the ghosting behind the car that is always common with the FSR2 mods yet glossed over DLSS/XeSS ghosting. The agenda is clear.

From my testing FSR2 does nto have any othe rghosting apart from behind the car, even mentioned it in the previous post.
 
Last edited:
I see you picked the ghosting behind the car that is always common with the FSR2 mods yet glossed over DLSS/XeSS ghosting. The agenda is clear.

From my testing FSR2 does nto have any othe rghosting apart from behind the car, even mentioned it in the previous post.

You stated right in your post that dlss and xess have ghosting on car tail lights yet when you selected fsr 2, there was none.... Hence why I created the above video.

I did not see your edit where you added the extra info hence why your updated post is not in my OP to you here:

9cZjgqa.png


There is still ghosting in other areas even with fsr, I'll upload another video sure.
 
Last edited:
@fs123

Where exactly is a good spot to test ghosting that isn't resigned to cars? As I'm having a hard time finding anything which is "noticeable" with any of the upscaling tech and FG tbh. Found this guy running with a bag which exhibitis some trailing, more noticeable with FSR.


Either way, main point is I'm not really seeing anything which is screaming out to me that amd are optimising their frame gen to work better with their upscaling tech over others, maybe in the official integration that may be the case but alas, we can't use any other upscaling tech with their frame gen in such scenarios so will never know unless that changes....

With the mod DLSS and XeSS do to have ghosting, especially on the car tail lights. However, when I selected FSR2 there was almost none (apart from behind the car) which indicates that AMD has indeed optimized FSR3 for their own upscaler. It is obvious they will not optimize for other upscalers. FSR2 does have a lot more shimmer than the native implementation though.
 
Last edited:
@fs123

Where exactly is a good spot to test ghosting that isn't resigned to cars? As I'm having a hard time finding anything which is "noticeable" with any of the upscaling tech and FG tbh. Found this guy running with a bag which exhibitis some trailing, more noticeable with FSR.


Either way, main point is I'm not really seeing anything which is screaming out to me that amd are optimising their frame gen to work better with their upscaling tech over others, maybe in the official integration that may be the case but alas, we can't use any other upscaling tech with their frame gen in such scenarios so will never know unless that changes....

I couldn't see any ghosting on any of the settings in your video.
I just checked again and it looks like the main cause of ghosting is from Ray Reconstruction. When moving side to side or back and forth looking at a fence or overhead powerlines you can see trails with DLSS +RR. Disable it and ghosting is gone and it appears around the same as FSR2 or XeSS.

Overall I reckon FSR3 FG is good enough for anyone who needs to gain some fps in heavier games. CB 2077 PT mode needs it for sure on a 3080.
 
Last edited:
Both are just as bad as each other, neither company is good and it's foolish to think amd are the good guys, they aren't, you buy what fits your needs and budget at the time, don't let morals get in the way as neither company has your best interests at heart.

This.
 
I'm sceptical about the differences in frame generation between FSR3 and DLSS3, and I'm not sure if people are testing at native resolution (or with some upscaling).

I also think FSR3 is probably not going to work quite as well on Nvidia hardware, not really a problem for games with DLSS3 support.

Either way, more cards and games are getting the ability to use frame generation (which can't be bad), which is still a pretty new technology.

There seems to be lots of pessimism regarding AFMF, considering it's still a beta (and not due for full release until some point in 2024). I think when it becomes more mature, game developers will want to make use of it in console games, or maybe Microsoft and Sony will give players an option to enable /disable this feature globally.

Antilag+ (removed to prevent issues with bans in some MP games) is still due for release also, which I assume will be of some use. But, I don't see AFMF being widely used by gamers unless they can avoid switching it off dynamically, with any fast camera motions.
 
Last edited:
Regarding AMD's cards, I think they are still offering the best cost per frame at MSRP, for their latest generation:

performance-per-msrp-1920-1080.png
performance-per-msrp-2560-1440.png


Technically, the RTX 4060 is a bit ahead in performance per £, but there's no FE version of this card... Also, the performance at 1080p isn't really ideal in some of the latest games, like Alan Wake II. Both the TI and non TI have just 8Gb of VRAM...

The RTX 4070 has a better spec (including 12GB of VRAM), but the MSRP is £569 in the UK, which likely means this card is never gonna be particularly affordable. Still, maybe some of these cards will be sold for £500 next year.
 
Last edited:
Back
Top Bottom