Question FSR 3.0 - AMD's Answer to DLSS 3.0 has landed

[uzzi38]

FSR 3 is not any better than DLSS 3, so they are not adding value, but merely not losing value by having it.

I'd argue the frame generation portion is. In gameplay image quality differences are nigh on impossible to notice, while FSR3 framegen has significantly lower compute times than DLSS3 framegen.

AMD really needs to step up their game on the upscaling portion as it's lagging behind by a decent amount, but the framegen portion is very solid.

 

[Aapje]

I'd argue the frame generation portion is. In gameplay image quality differences are nigh on impossible to notice, while FSR3 framegen has significantly lower compute times than DLSS3 framegen.

AMD really needs to step up their game on the upscaling portion as it's lagging behind by a decent amount, but the framegen portion is very solid.

Does this distinction matter when you can't use FSR 3 with DLSS 3?

 

[SolidQ]

AMD really needs to step up their game on the upscaling portion as it's lagging behind by a decent amount

More interesting is this

AMD FSR is the building block for Apple's MetalFX upscaling tech — the app's legal info references the usage of AMD FSR

This makes much sense, considering that AMD FSR has been open source, and MetalFX yields comparable results.

www.tomshardware.com

 

[Mopetar]

I'd argue the frame generation portion is. In gameplay image quality differences are nigh on impossible to notice, while FSR3 framegen has significantly lower compute times than DLSS3 framegen.

Arguments over which company has the best fake frames is like arguing over who has the purest strain of herpes.

 

[psolord]

I did a new test, with fsr3+FG on my old 2500k+6600 system, in Immortals of Aveum. Very interesting, as the settings I used were downright wrong on purpose.

The game has a numeric weighing of each setting and also assigns a numeric capacity of your graphics card. The rx6600 has a capacity of 394 points and the settings I used have a weight of 690 points.

I start my run with no FG and I show the difference afterwards at around 1:00. It is indeed knight and day. The game goes from 40-50 to 60 vsynced.

Also I am using fsr3 quality here, so essentially the resolution is 720p. Visually, it's a pretty good result if I may add. The game also runs at 720p internally on the big Gen9 consoles, as a point of reference, if you need to complain...This is a super heavy UE5 game and come to think of it, a mere 6600 can now provide an even better image quality, thanks to fsr3.

Here's the vid, hobbyist non monetized as always.

Also keep in mind, there's a fsr3 fg mod that has been leaked, that can be used in a number of games that support dlss3. Instead of dlss3 you use fsr3+fg on amd cards and it even works for the likes of rx580, gtx 1650 and so on. Youtube is currently getting filled with these comparisons and results are great. Here are some examples

https://www.youtube.com/results?search_query=fsr3+mod+test

I have these files and I intend to test some heavily cpu and/or gpu limited games, on this 2500k+6600 system. Ratchet, Jedi and Starfield will be very interesting...

I think amd is marketing fsr3+fg wrong. Instead of advertising high framerates being even higher, they should put more weight of games becoming playable from unplayable.

I am now going to post on the 8gb not enough thread, that with these mods, you are essentially downloading vram from the internet and see people there lose their minds. xD Nuh, I am good guy.


Happy New Year everybody.

 

[Heartbreaker]

I start my run with no FG and I show the difference afterwards at around 1:00. It is indeed knight and day. The game goes from 40-50 to 60 vsynced.

That tends to be a poor outcome, with significantly worse input lag a jittery feel from what most testers report. You are basically adding another ~25 ms of lag to poor frame rate, which is a LOT.

FG has a narrow range of usability for most people, that starts with native frame rate already above 60 FPS.

IOW it isn't really useful for improving low frame rates. It's mainly useful for motion smoothing on already decent frame rates.

 

[Aapje]

Exactly. And that is exactly why frame generation is way less useful than temporal upscaling. The latter actually allows you to run games that you couldn't run otherwise, but the former doesn't.

 

[psolord]

I know what I felt guys and it even shows in the video. The game went from a jittery mess to a playable experience, 25ms or not.

You can easily test it on your systems with forspoken demo. Only Aveum is even better.

 

[GodisanAtheist]

That thumbnail of Avenum makes it look like a successor to Blood or Hexen or something...

 

[Heartbreaker]

I know what I felt guys and it even shows in the video. The game went from a jittery mess to a playable experience, 25ms or not.

You can easily test it on your systems with forspoken demo. Only Aveum is even better.

Like many things. Some people are very sensitive to lag, while others aren't.

There is no debate that you are adding around 25ms of additional lag, which is a LOT extra lag for many people.

Are you playing with a controller? That tends to mask lag to a certain degree.

Other reviewers indicate further problems using it at low frame rate FG with Vsync. Where it can make things even worse by pausing real frames in favor of fake frames.

I have an RTX 4070 so I can run both NVidia and AMD FG, but have ZERO interest in doing so, since I don't have a high refresh monitor and I can run everything at smooth 60 FPS native.

 

[Hitman928]

Like many things. Some people are very sensitive to lag, while others aren't.

There is no debate that you are adding around 25ms of additional lag, which is a LOT extra lag for many people.

Are you playing with a controller? That tends to mask lag to a certain degree.

Other reviewers indicate further problems using it at low frame rate FG with Vsync. Where it can make things even worse by pausing real frames in favor of fake frames.

I have an RTX 4070 so I can run both NVidia and AMD FG, but have ZERO interest in doing so, since I don't have a high refresh monitor and I can run everything at smooth 60 FPS native.

I downloaded the demo and a quick test impression for me meant keeping FG off. I tried (with an RX6800) 4K native with highest settings which gave me around 30 fps. Turning on FG changed nothing in terms of how the game felt or looked to me despite the frame rate jumping up to about 45 fps. Granted, I have a VRR monitor that can stay synched below 30 fps. At 4K with FSR Quality I was getting around 45 - 50 fps. Turning on FG again seemed to make no difference in terms of feel or visuals despite the fps now being close to 80 fps. I didn't really feel an increase in latency but I was also just at the very beginning of the game mostly just running around and doing the in plot tutorial on how to attack. I wasn't doing anything where response time was actually important. So in this very short test, FG really made no noticeable difference.

With that said, if I didn't have a VRR monitor, going from 45 - 50 fps to a locked 60 fps with Vsync might make FG preferable to get rid of the screen tearing, even with an added latency. This would improve the visual and probably overall experience to me significantly as I tend to be pretty sensitive to screen tearing in most games. I think getting a VRR monitor is probably the better solution, but if you don't have one, I could see FG being useful in getting the last 10 - 15 fps in performance to hit 60 fps. I think once you get close to 30 fps though, the latency just becomes too much and you just need to turn down settings to improve the experience.

 

[Heartbreaker]

With that said, if I didn't have a VRR monitor, going from 45 - 50 fps to a locked 60 fps with Vsync might make FG preferable to get rid of the screen tearing, even with an added latency.

Everything I've seen, says it really doesn't work well with Vsync. You need VRR or it actually introduces tearing, or frame pacing issues.

 

[psolord]

Like many things. Some people are very sensitive to lag, while others aren't.

There is no debate that you are adding around 25ms of additional lag, which is a LOT extra lag for many people.

Are you playing with a controller? That tends to mask lag to a certain degree.

Other reviewers indicate further problems using it at low frame rate FG with Vsync. Where it can make things even worse by pausing real frames in favor of fake frames.

I have an RTX 4070 so I can run both NVidia and AMD FG, but have ZERO interest in doing so, since I don't have a high refresh monitor and I can run everything at smooth 60 FPS native.

I have a 4070ti and have never used dlss3 because it sucks. I too prefer using correct settings, in order to get as close to 60fps vsynced as possible.

That is a totally different matter however, to what we are discussing here. Meaning, a problematic game, that stays problematic even with fsr3 upscaling. At that point FG+vsync makes it better.

I use kb+mouse always.

 

[psolord]

Everything I've seen, says it really doesn't work with Vsync. You need VRR or it actually introduces tearing.

I was using vsync with no problem in Aveum above.

Tim here revisits fsr3. He says that with vrr displays the gain is not as good as with steady vsynced monitors.

Also yes, it improved Avatar too, but introduces UI artifacts, because apparently the devs, should live the UI out of the fsr3 pipeline, but didn't get the memo.

 

[Hitman928]

Tim here revisits fsr3. He says that with vrr displays the gain is not as good as with steady vsynced monitors.

That was only with the early version of FSR FG which had issues with VRR monitors, the updated version works fine with VRR monitors and using FG with a VRR monitor is the preferred way now (if you are going to use FG). I didn't see the issues he mentioned with FG and Aveum in my quick play through but I'll try it again, it may be that I was just in a good frame rate range compared to my monitors max refresh rate that I didn't experience the issue.

 

[Hitman928]

Everything I've seen, says it really doesn't work well with Vsync. You need VRR or it actually introduces tearing, or frame pacing issues.

I haven't seen this but I honestly haven't been paying much attention to FG tests due to lack of interest. I'll try it myself and see if I get the same thing.

 

[Ranulf]

FSR3 exposes the ugly side of PC gamers or something. Fake frames vs old cards new again I guess.

AMD FSR 3.0 has exposed the ugly truth about most PC gamers

From comments like "I don't want fake frames", we went, in the blink of an eye, to "My RTX2060 lives again, FSR 3.0 doubled my performance".

www.dsogaming.com

 

[gdansk]

I don't want fake frames as a justifying feature to spend hundreds of dollars on a mediocre GPU. I don't want it used in benchmarks that lie about being 3x faster than last gen.

But I guess if it makes a mediocre GPU I already have a bit more fluid for free then OK cool.

 

[GodisanAtheist]

I don't want fake frames as a justifying feature to spend hundreds of dollars on a mediocre GPU. I don't want it used in benchmarks that lie about being 3x faster than last gen.

But I guess if it makes a mediocre GPU I already have a bit more fluid for free then OK cool.

-100%.

FSR 2.0 let my 980Ti Play Hogwarts Legacy at an acceptable framerate, I have no issues with that.

FSR/DLSS/XeSS being essentially required to play a game like Alan Wake 2 on even top of the line $1000+ GPUs? That is outrageous.

 

[nurturedhate]

-100%.

FSR 2.0 let my 980Ti Play Hogwarts Legacy at an acceptable framerate, I have no issues with that.

FSR/DLSS/XeSS being essentially required to play a game like Alan Wake 2 on even top of the line $1000+ GPUs? That is outrageous.

This

 

[DAPUNISHER]

But I guess if it makes a mediocre GPU I already have a bit more fluid for free then OK cool.

Yeah, I mean if you have a scenario like a broke kid with a hand me down 1050 getting a better experience, why not?

 

[Aapje]

FSR3 exposes the ugly side of PC gamers or something. Fake frames vs old cards new again I guess.

AMD FSR 3.0 has exposed the ugly truth about most PC gamers

From comments like "I don't want fake frames", we went, in the blink of an eye, to "My RTX2060 lives again, FSR 3.0 doubled my performance".

www.dsogaming.com

FSR3 exposes the ugly side of the media, creating a narrative out of thin air by selectively quoting some people to create this false narrative where all gamers had the same opinion beforehand and then suddenly all adopted the opposite opinion.

The lesson here is that this DSO Gaming website is a rag that produces clickbait.

 

[Heartbreaker]

Yeah, I mean if you have a scenario like a broke kid with a hand me down 1050 getting a better experience, why not?

HUD is a jittery mess. I'd rather just disable more setting to have higher actual frame rate, and stable HUD.

 

[coercitiv]

The lesson here is that this DSO Gaming website is a rag that produces clickbait.

Take a quick look at the comment section in that article: the author and owner of the site is getting the full treatment from his own readers.

 

[Fanatical Meat]

Been out of the loop on this is it software that needs to be added or just something the game has built into as a setting?