Problem with DLSS frame gen is that Nvidia use it as a performance metrics to compare it with previous gen in their marketing materials for the 40 series.

A game needs to reach 60 FPS by default, then FG is cool for singleplayer. In "competitive" multiplayer however "real" frames are worth more, because they actually respond to input. Generated frames are only visual, they don't have anything to do with input. Edit: added the word competitive

FG should be an afterthought not a requirement.

A benefit from Frame Generation I don't think I've ever seem mentioned is it allows for games with framerate dependent things like Skyrim's physics system to rum "over" the intended framerate without breaking those systems.

I have 3080, I originally tested FSR3's FrameGen in Forspoken demo and I really didn't like it. It felt a bit stuttery, uneven and unclear. BUT then I followed your last video about the FSRxDLSS FrameGen mod and modded in the FSR3's Framegen into Starfield and... holy hell, not only it worked, but it actually performs amazingly there, and looks great coupled with DLSS. It's super smooth and I don't notice any artifacts and issues. For a single player game like that it's awesome and it boosts the FPS nicely in the unoptimized places, making the native framerate drops in cities basically invisible. A bit higher input lag, sure, but then again, I don't need super reflexes in Starfield ;-)

There was a negative reaction before because people couldn't afford it. And its positive now because fsr 3 has allowed everyone to have it.

The latency is around 1 frame time more with frame gen, so if the native fps is high, the frame time is low so the latency is relatively low. I think if you use frame gen and lock the max fps, it can save power so it maybe a solution for handheld gaming. If the native fps is 40, the latency per frame is around 25ms, it is a bit lower than cloud gaming but the effect is similar, some people feel ok some not, but this is very good for us to choice this feature.

I think the "avg" user gets into the game and doesn't really notice as long as that base FPS is snappy, most times it's only a thing when I'm not really into the game

Something similar to frame extrapolation is widely used in VR, it is called asynchronous space warp (ASW). The images in the headset have to follow head movement very quickly to avoid motion sickness. For many people native, locked 60 FPS is too slow in VR. The problem usually disappears at ~80-90 FPS. ASW helps with that. If you move your head, it will take the last frame, shift, rotate and zoom it according to your head movement and apply some perspective correction -magic- math and use that as the next frame. This is much quicker than a traditional render. It is very common to play a demanding game like MS flight sim in VR at a locked 45 FPS and use ASW to bring that to 90. (Or turn 60 FPS into 120 if your computer can handle it). The VR headset for the PS4 had a "controller box" that even contained special hardware for that task. We don't have that for PCs or the PS5 because modern GPUs can do it with ease. This will not smooth out motion of other players. It will also not speed up hand movement to image latency, there is no advantage in games like Beat Saber. But head movement to image latency is greatly reduced. It seems logical to use something similar to reduce mouse to screen lag. Looking around with a mouse on a flat screen game is very similar to looking around in a VR game with your head. If you push the button, the bullet won't come out faster. But it would greatly reduce the "rubber band effect" that you get when looking around with a mouse at 30 FPS.

Biggest issues are that they are dependent on implementation to be good or bad. Next to having a varied of each of the tech being available per game... Yes, developers can put minimal time into it to put it in. However making it look and perform well takes time and effort. DLSS in a lot of games suck on my 3080 but when I tried FSR3 on Avatar it looked amazingly because I feel like that game is really optimized to show off what FSR3 can look and feel like, and you do get more frames and played awesomely. TAA being enforced because of it, kind of sucks for those sensitive to blur, like my own eyes. Which is often forgotten that some people may not be able to use it. Generally speaking do think because little amount of people got access to it and it being limited to only one of the latest and MOST expensive generation of graphics cards made a lot of people upset because of course they would like this as well. I still would like more raw pixels, no TAA being enforced, being generated like old school traditional rasterzation performance. Games already became blurrier because of deferred rendering engines vs forward rendering engines we had before. Personally I think RT is really cool tech, but the offset performance for that visual trade is still too much. I don't want to discredit these technologies but they in essence are bandaids to overcome the lack of performance we can get out of games. I don't mind if a game looks a bit worse to make it play well, it is a game. Not a movie. and I feel like we lost that touch by making it too much movie-like experience.

I think Intel has the best idea. Instead of interpolating, they extrapolate frames. Solves the latency issue. It has it own set of issues, but the direction makes sense.

20:00. I love the idea of leveraging software to improve performance where additional optimizations aren't possible but it's hard to see 3x, 4x, and 5x performance amplifiers which have traditionally relied on hardware upgrades. I've already seen a bunch of comments talking about there is no longer any need to upgrade their GPUs soon because of this open source upscale friendly frame generation mod. Seems like an advanced version of this technology would eat into GPU sales.

upgraded from a 3060ti to a 4070 for frame gen, 3 months later and I never use it. Played lots of game with it too recently but it has too many compatibility issues with things like HDR, G/V Sync Etc. and different games require different setting to make it usable and a high frame rate to begin with. NEEDS WORK.

I agree with the sentiment, manufacturers are going to provide less card fkr the dollar, relying on FG to boost to "appropriate" performance, and game devs will falter to working less to make the game run acceptable without it.

I tried FSR3 in CP2077 and AW2 and I'm not really sure if I would use them. I like high framerates because of the 'snappy' feeling and FG doesnt give that. If I have a base framerate high enough that the input lag isnt noticeable, then I feel its also high enough that FG is kinda pointless. 120 real frames already looks and feels great. I dont need to get marginally better smoothness at the cost of input lag and more artifacts.

Dude, i cannot thank you enough for explaining FG with the FSR3 works. I love Cyberpunk and using that FSR3 mods was magic, i can run the game cranked up to the max and it plays well :D. I then tried it on The Witcher 3 cranked up to 11 and oh my god was it beautiful, i always played it on dx11 because i could run well the raytracing version but now its butter smooth :O

Very rational video Daniel. You hit everything I like and dislike about frame generation. It is a net positive, and this is just the beginning. Drivers and frame generation will develop and mature.

You are not seeing the whole picture. Is FG a good tech? Kinda. It's not bad in itself. However release of DLSS and FSR excused a lot of publishers to push unfinished and unoptimized games. This is exactly the situation that FG will cause in the future and we will be told: "Just turn frame generation ON". Already there are games that turn upscaling ON by default. The games look worse, we are getting lower frames overall and we are being told to upgrade.

I find the point at the end (looking toward the future) interesting. But the biggest question I have is input lag. Even in solo games, I notice that more than raw visual framerate most of the time (but that's going to vary person-to-person). To use your example of a 1,000Hz monitor, if we could get to the 557th frame of that (a generated one) and it could go "Wait, the player just hit a button!", and it could react to that, I'm 100% on-board. Even if the game didn't display the difference right at that moment. Think of the whole "predictive movement" thing in online games, where it assumes another player continues their same direction, then "goes back" and changes it when your client hears the other player changed direction a quarter-second ago. What I'm picturing is sorta like that, on much shorter timescales. Like even if the next 10-20ms of frames don't *show* my input, the game engine registers them and the effects of my keypress shows up after that. The idea is that at that point, there's no reason to really *lock* input-handling to spitting out game-engine frames. From a programming standpoint, you would have a thread that spits out frames, and a thread that process inputs. These happen asynchronously, and each just goes as fast as it can. 🙂

"Renderframe" vs "Interpolationframe" may be the most suitable/descriptive terms. 💙

I tried the DLSSG to FSR3 mod with a bunch of games and it works amazingly well. Some games got some HUD issues tho, most notably Ratchet and Clank, but the majority of them it works really good.

The thing is that once the generated frames aren't noticeably garbled, and it's harder to notice in higher frames because those artifacts are generally temporary, the main argument is that it degrades input latency. However games that have DLSS3 have natively forced Reflex onto them which improves that. In previous tests it was proven that input delay with FG & Reflex is either better or as good as no FG and no Reflex. And let's be honest, none of those games would have Reflex if it wasn't due to DLSS3 being in those. So it's a net positive either way. If someone really prefers better input latency, they can skip it. I'll personally take fluid motion instead until I realize I'm not doing as good as I want to do. But yeah, games shouldn't arrive at a point where we need framegen in recommended specs for 30FPS.

"FG frames are worse than real high fps, but better than real low fps. " From a user of FG since day one!

Though not quite the same as DLSSFG or FSR3, I've been using FMF extensively on my 7900XTX, and its been an absolute godsend. Its been particularly great in games that are either capped at 60 due to engine restraints or unable to reach a higher FPS due to bottlenecking. I've been playing a ton of games on the Ryujinx Switch Emulator and while for most games you can use a mod to unlock the framerates from 30 to 60, there is obviously no way to unlock the framerate even further to say 120. I have to say, playing Fire Emblem Three Houses at 120FPS makes the game feel unbelievably smooth -even if the physical latency doesn't change for the better. I've also used it quite a bit in my heavily modded Skyrim that has a heavy ENB on top and it does wonders there too. It was also great for me in Starfield where I was CPU bottlenecked (feel like that's more the game than my CPU though.) I understand a lot of the negativity surrounding these technologies, particularly around the marketing and over-reliance on them; but honestly I'm a huge fan of this frame-gen tech and can't wait to see how it evolves further.

Personally I rather not use an upscaler or a frame generation, though I understand that there are situations in which frame generation and upscaling works better and I guess for those it is ok. Though for me if I can hit my good enough performance at a target fidelity without using either then I won't use them.

honestly for me, for those with VRR Displays. I say it's a overall improvement as long as you're above 30fps, play on controller, and the game's input latency is good on controller. For me I have a RTX 3080-12GB, so in Cyberpunk for example I can hit ~40-50FPS at the lowest at 4K Ultra Performance DLSS Mode with Path Tracing and Ray Reconstruction Engaged. So FSR-FG perceptually gives me 80-100FPS 98% of the time, and a bigger boon that I don't think is considered as much for FG, it helps act as Software Low Framerate compensation (but better), as while my TV would experience tearing or stutter if I dropped bellow 40FPS (As it's not a Freesync Premium TV, heck, I'm already modding it to let it VRR down to 40), with FG I wouldn't experience tearing/stutter as long as it works with VRR. Which the Public release of FSR3 does support FSR3, and in the modded version with Cyberpunk, it produces a stellare experience, so good in fact I could swtich from 4K UP (720p IR) to 1662p Performance Mode (831p IR) with NIS to 4K and get overall similar performance and still retaining above 30FPS at the absolute lowest internally. And considering how Cyberpunk's Path Tracer and Ray Reconstruction scale to resolution internally, that ~110p radically improves image quality despite the reconstruction output being lower before the less sophisticated NIS takes over to take it to 4K

CONGRATS on 169K ---- We have multiple Frames generated already -- VR does 22.5 and 30 to native 90 so 3 and 2 INTER frames generated - Motion Reprojection in OpenXR --

I'm curious if the artifacts in AW2 were due to motion blur + FG. Looks like its on. perhaps if you disabled motion blur it'd be a better experience for you.

I’m excited for the future of frame interpolation and extrapolation. Once they nail the quality of these frames it’s going to be exciting to see them figure out ways to generate potentially multiple fake frames between real frames and get some real life levels of smoothness. I’ve only used it in Avatar so far and it’s incredible to me

I use frame generation all the time bros. Its on my TV, and its what makes quality mode on a console bearable. The fact, though, that we are using this on our PCs is proof that 2012 will never return. Moore's law is not just dead, bros. It's extinct.

I'm happy with locking single player games to 60fps. No need to worry about fake frames.

Do you think input latency issue will somehow be fixed or at least be reduced in the future?

Frame generation from AMD's fsr3 is a game changer for rtx 3060 gamer like me.

I think a key term to use here instead of "picture smoothness" is "motion clarity" (borrowing from high refresh rate monitor discussion). All else being equal, it will be easier to notice or track fast moving elements (enemies or items when moving the camera quickly) with well implemented frame gen. I completely agree with the marketing point, I was initially very negative about it when they marketed it as just double performance

I tried to use the FSR 3 mod with skyrim (and other optimization mods) to see how high my framerate can get but sadly it doesn't work it yet because skyrim is directx11 and not 12. However it does work with Elden Ring and I was able to get 120 fps on my 2080 (although the HUD was all messed up when moving)

DLSS/FSR and FG should be complimentary. A feature to help consumers get some extra longevity from their hw. But now it’s just mandatory to use it even with the latest and most expensive hw. That is the problem. You need them to play even crappy looking games like starfield.

Extrapolating is not that new as a technique. Every wife can extrapolate every single thing you did wrong since Middle Age to generate frames against you during a discussion, even if you are right.

One of the problems is people usually see reviews/metrics on KZfaq before they experience the technologies themselves. It's a lot harder to notice some of these issues if you haven't already been told by someone that they are there and that you should be bothered by them.

Will have to judge this on game by game basis, maybe even setting by setting tbh. Found a few individual settings which cause visual issues with AMD FG but it's mostly fine with them turned off.

i have more fps but not smooth with vsync on, any idea to fix it

Not had a chance yet to test frame generation, only have a 30 series. But I get the feel it will be like DLSS as you say. It is far better to not have to use it but in the right situation the gains are worth the downsides.

TIMESTAMPS 0:00 Introduction: Frame Generation getting better reception then at launch 0:51 Why was the reception of Frame Generation so negative when it first launched 2:52 More people been able to use Frame Generation and been happy with the results 5:05 The downsides of Frame Generation 7:42 The benefits of Frame Generation 9:35 Personal experience of Frame Generation 13:07 DLSS Frame Generation better than FSR 3 for frame pacing, FSR 3 is still good 17:29 Frame Generation is not pointless 18:05 The future of Frame Generation 23:20 Conclusion/Final Thoughts P.S.: If you feel that these timestamps misrepresent what is presented in this video, let me know!

Frame gen is cool tech that I like using depending on the game like you said. If I care about the quality of my human performance in a game, no I don’t use it and shoot for typically at least 100 native frames. Even if input lag was null, I’d still likely keep it off unless the smoothness felt like a competitive advantage, because I don’t want to introduce hallucinated feedback (looks like I hit someone but didn’t). Maybe my opinion on that would change with usage on a monitor that’s 240hz or higher though. But even with AAA games, some of the artifacts are really nasty. Like any area behind the transparent hud in avatar. That alone makes it unusable to me, it looks like localized screen tearing due to the different update rates.

For fsr3 u just need a cap, doesnt need to be at your monitors max refresh rate. For amd users just cap your framerate (should be low enough that you are locked at it) from the amd overlay. Has perfect frame pacing

I've wanted KZfaq to support higher framerates for awhile and I'm really curious to hear the opinions of other creators like Daniel, who would be using it for practical purposes in these reviews. Thanks for the content as usual!

Most big studio are going the 'RayTraced GI, Shadow, Reflection' era. Avatar is a nice example. It runs on none RT hardware but it's super slow, but it runs. Snowdrop V2 at ubisoft is going to be in a LOT of games going forward. Lots of UE5 dev will use LUMENS. It's a win win for both of us. It's faster for them to make the video game and it look better to us!

Playing Cyberpunk 2077 with the modded fsr3 frame-gen bumped my fps from ~60 to ~120. Dlss quality, high settings. It feels smooth. Very, very smooth. Even on kbm (though I prefer to play with Dualsense, the new adaptive thingy makes the game so much more immersive), it feels like magic to me.

Thanks for your tempered thoughts, much appreciated. I still haven't seen FG in action, but I am very curious about this technology. One concern is that I don't really see much difference with anything above 60Hz (except for multiplayer shooters) - as you say, lag perception is very subjective. So if FG gives best results on base fps above 60, then it may not make much difference for me, personally. Anyways, I look forward to being able to afford one of the newer cards and finally experience FG (along with other cool new things, PT being the most attractive).

I've been playing cyberpunk with path tracing and fsr frame gen mod on a RTX 3060 achieving around 75fps. It doesn't feel too bad for me and I really prefer motion fluidity sacrificing some latency in a single player game like this.

Personally frame interpolation is still a negative for me. lower input latency is what i want from higher framerates and visual clarity is just a bonus. Intels extrapolation technique sounds more interesting to me. I hope its good and AMD and nvidia will build something similar if it is.

If you play single player games its awesome. Set your settings to where you're in the 40-50fps running something like native or DLSS Quality then turn it on. Great experience. For multiplayer head clickers... not so much.

How do i use fsr 3? I got amd graphics card and only appears fsr2

I play on linux with a Ryzen 7700 and a rx6600xt with a 1440p 165fps screen with freesync. I've not yet seen much frame generation but on CP2077, upscaling, in auto mode, gives me a solid 60+ fps. Look forward to FSR3 (frame gen) in CP2077 and other games. Its a tool, if gives me a better experience I'll use it, if not I will not. So far, FSR 2.x has usually helped - suspect FSR 3.x will be similar (I have used FSR 3 in Forespoken, it was ok but I consider it a 'release candidate' not the product that FSR3 will end up being). Bottom line here is that FSR is improving my game play and is letting me stretch my GPU while waiting for better $/frame...

Ppl did not hate technologies of frame generation, they hated Nvidia for bad generation performance upgrade + bad pricing and covering this problem with DLSS 3. "Yes, rtx 4070 has got the same performance as 3070ti and costs +$100, but if you turn on DLSS 3, you may see the difference..."

I don’t think extrapolation needs any doubts right? Isn’t that how the latest VR headset (Quest?) is pumping out more performance, where it matters even more and does it successfully?

I have a video idea, if you’re interested. I’m new to pc, and I’m confused by the jumble of technologies (acronyms and abbreviations) available to improve your graphics. I do try them out and attempt to find what looks best, but don’t really have a frame of reference. Anyway, the video idea or series is: using real a game example, show how to lock your game to the monitors refresh rate, like you mention in this video. After that video you could do a series of videos in that same vein. Highlighting each technology, and how to optimize it in specific games.

The only people who hated Frame Generation are people who couldn't experience it. Just like DLSS 1.0 back in 2018

The negativity about FG is 90% NVIDIA's fault: - The lock of FG to 40 series is probably just a sales tactic, since they sat on a literal mountain of 30 series chips at the time of 40 series launch and needed a tangible gimme for gamers and it's not the first time they claimed something was hardware locked only for people to dig and find an easy way to make it work on their excluded cards that "lacked hardware support", most recent example: NVIDIA Broadcast launch version working on 1060 cards when they claimed it used RTX specific cores. - 40 cards being so overpriced at launch, so the 30 card mountain would eventually be sold off, essentially making FG a $500+ feature that would be really useful on a $250 card. - Scummy NVIDIA graphs and naming that bundled DLSS3 & FG into one thing, while it is two clearly two separate components as people digging have found out (you can combine DLSS with AMD FG for example) - Scummy NVIDIA marketing with their DLSS3=FG slides comparison vs older card comparison for a feature that just~20 games out of the box support a year later. The other 10% is AMD's fault with the first few games having a crappy launch implementation, essentially pre-murdering their hype with the first two games being wonky due to using outdated FSR versions.

i love xess in cyberpunk but i have a 30-40 fps difference in 2.5k in vr, thats the difference of motionsickness to smooth gameplay in vr. Plus xess only looks good when you stand still, its super blurry at 45 fps.

Why not get the Avermedia Live Gamer ULTRA 2.1 to record the 4K 120fps footage directly? Just a question not a critique. I enjoy your content.

One thing I *CAN* say: I *NEVER* want, nor will want, upscaling or frame generation included in pure hardware tests (e.g. GPU reviews), because that will make it harder to have an apples-to-apples comparison when selecting my hardware. A separate test, either in a separate section or a secondary review, is not a bad idea - but I want to know how much pure render power I'm buying, not how well it performs at a lower resolution than tested or how many additional frames it can construct without actually rendering.

If the cards had the hardware to render it natively then we wouldn't "need" frame generation at all. They cut back on the hardware and are trying to push the "feature" THATS my beef with FG

Nice explanation. For me base ~50fps is enough in single player games to go frame-gen ON. But there is also one more caveat with DLSS3, which is: switching between static menu screen back and the gameplay back and forth brings nasty stutter for some time after transition. The same happens in CDPR path-traced titles, but also in Alan Wake 2. If you have a single player gamer reaching constantly to your inventory screen this is till super distracting with DLSS3, even more then some slight motion artifacts during regular gameplay, at least for me.

For the Alan Wake 2 example can you try the comparison with motion blur toggled off. Even without those artifacts I would play with motion blur off when using frame gen because its applying the blur in the context of a lower frame rate, so the blurring beyond the motion of the camera.

I'm just sick of people calling frame generation a performance feature. It's not a performance feature; it's a visual smoothness feature.

I like the tech, I despise the sales pitch and marketing angle. If they had named it "visual smoothening" or "motion blur+" then I wouldnt have a problem with it. It being marketed as "fps gain" or "free performance uplift" should be fking illegal.........

There is one thing i don't understand. Why gets v-sync disabled when using FG ? I'm using a g-sync display at 165hz. Everything i need to enable V-sync in the driver for that particular game other i get screen tearing. I thought that FG was supposed to be used with v-sync. That was happening in 2 games, Starfield and Ark Survival Ascended.

I got a 7800 XT few days a go and tried FSR3 in Avatar game, fps went from 80 to 120 but it didnt feel good at all, felt really stuttery when looking around. Idk if all my setting were correct for FSR3 though, have to dabble more with it I think.

I've only used AFMF so far on my 6900xt and I absolutely love it. cant wait for it to be built in to the driver instead of having to use beta drivers

I don't want to HAVE TO use FG when i buy new hardware, i want to BE ABLE to use FG with my existing hardware to make them longer lasting.

I also feel like this where we're going and I'm not excited about it at all. Every kind of motion interpolation I've ever tried felt weird and uncomfortable to me, even without taking into account an increased latency. I also feel the exact opposite regarding the controller vs mouse latency perception. For me latency is more disrupting when playing with a controller. I think that it might be due to the fact that it relies more on the temporal aspect of input rather than spatial (to compensate for the obvious ramifications of the thumbstick design), hence being more dependent on timing. It's interesting that we have such a difference perception here.

I wonder if vrr via software will be possible as well

frame generation is pretty cool, but as a person who gets motion sick in first person games i am still on the fence about this technology as my source of motion sickness often comes from input latency rather than smooth movements. Because frame generation increases input latency it is hard for me to weigh the tradeoffs of playing a first person game at a lower framerate for longer rather than a higher framerate but experience motion sickness sooner.

After trying FG myself i felt scammed. Basically you need 60 fps baseline or else it will feel really floaty and input lag will be worse than native framerate. (imagine it was 30 it will have way more input lag than if you just rendered 30) So it's only useful if you want to run a game that you already can run at 60 fps at 120+ fps... which is just so weird? because the only games that i play at higher than 60 framerates are twitchy multiplayer shooters and similar competitive stuff, which don't need FG because they are games that already run at 200 fps by default!! I play all my singleplayer/story/graphics heavy games at 60 fps and thats plenty for an amazing experience, interpolating frames to boost that to 120+ in my eyes is nvidia creating FOMO for a small gain ( motion becomes more fluid if you have high Hz monitor), so def not worth buying 4000 series gpu. I also think path tracing whole scenes and GI for games that don't have day nigh cycles is ridiculous and borderline dishonest from devs to seemingly make people think they need to buy these expensive gpus. I am a software engineer and really into game engines. All you need for most games (static scenes and static light sources) is raytraced reflections for curved surfaces, planar surfaces can be handled with planar reflections and its faster if you use correct LODs, everything else should be baked! Take Alan Wake 2 as an example, there is no reason for that game to have whole scene ray tracing. All it needed was RT reflections and maybe RT shadows as a future proof toggle. The enviroments are hand crafted static scenes!!

I have a gtx 1650 laptop is this good news or not ?

I slightly curious of the difference of using Nvida card with AMD card, but AMD only doing frame gen and how that compare vs doing frame gen on 1 card, I going assume its worse with 2 cards, but still curious

I have always loved all the ai tech DLSS is so good! And now frame gen aswell❤

I maintain the same stance on it I've had since FSR 3 was announced: DLSS 3 FG is not bad as a tech, it's bad because it has allowed Nvidia to massively upcharge customers on a bad card gen. Even if I cared very deeply about lag and image quality compromises that the tech may bring, it's not as though anyone is forcing me to use it. Everyone having an option to turn on FG and see if it makes sense for them is a net benefit; the way Nvidia implemented it was just scummy. You can argue about it setting a dangerous precedent, but honestly it's not targeted at a performance range that would allow developers to compensate for poor performance with it.

I just tried it and it's crazy how well it works on Spiderman Remastered for Steamdeck. Previously, I played on almost minimum settings and got like 40 FPS. Now I can almost play on highest settings and play on like 60 FPS, barely going down to like 55 FPS

I recently got a Meta Quest 3 and upscaling is used in vr games to improve frame rate. For VR the higher the frame rate the better for smoother game play. I wonder if FSR 3 with FG will help VR games or because of artifact's will make the game look bad in VR. There is no VR game I know off that uses frame gen out now but I would be interested in modding it in to see if it makes the game smoother in my Quest 3.

As long as it's being used in marketing (and being parroted by other people) as a "frame boost" or "performance boost" and is being used to say "Look at how much better THIS card is compared to LAST gen!" I will always be reluctant to actually care. The more I hear about frame interpolation the less I actually care to do it. Someone actually said "Digital Foundry put up a video of the interpolated frames and only those and the output was absolutely perfect, as good or better than the original" and I cringe at their hard work being subverted like this. You WILL eat the interpolated frames as performance boosts and you WILL like it, even though it sucks at a lot of stuff. The swing in attitude is 100% marketing shills.

6:35 I would consider them “fake” frames as they do not respond to input. Even better I think the tech should be called “image insertion” or something similar.

I am more excited for Intel's approach for Frame Generation, which they call ExtraSS? I think. But what different about their approach is that they don't HOLD 2 Frames to Generate a new Frame. But rather, from what I understood, they will try to predict what the Image should look like instead which from what they say will still have a Latency increase BUT will not be as noticeable as FSR 3 or DLSS 3's approach which HOLDS 2 frames and increases the Latency. Which, if true, would be in theory better than both NVIDIA and AMD. Because the Latency increase will not be as noticeable and the experience of Latency will feel like the Normal Frames you are getting while at the same time benefiting from the Generated Frames for high Refreshrate monitors.

Can anyone here explain me if i can use fsr3 fg in witcher 3 on my 6800xt if not why

I'm playing Cyperbunk 2077 path tracing with FSR FG. I'm just blown away with how smooth the game feels. I went from 45fps to 90fps. Latency and artifacts are noticeable but very tolerable. I genuinely lost complete interest in 4000 series GPUs now. AMD just added 10 years to my 3070.

I think unless it's gets to a point where you can use it at a base framerate of 30 and get 60 it isn't honestly that useful for most people.

FSR3 allowed me to play cyberpunk with path tracing on a 3080, and took me from an inconsistent 40 FPS to 70ish. I play with a controller so the 40 FPS (with reflex) latency is good enough for me, and the only problem was the visual smoothness, which fg solves. Nukem9's mod also works for other games so you should check it out. I think FG is great, but should NEVER be marketed as "performance"

Tried the dlss+fsr3 mod on spiderman remastered so far and my experience has been really good so far. Its not magic but it gives me a fair fps boost, especially noticable when jumping through the city really fast.

i find interpolation visibly jarring, and the interpolated frames arent responsive to peripheral input. i can see and feel the difference...it degrades my experience. its an unpleasant technology meant to impress casual gamers with 'higher number better' style logic, and i would rather have more raster horsepower and optimized driver packages than have AMD or nvidia focusing on these silly surface-level marketing gimmicks obviously only designed to increase mindshare for cards that are already selling themselves anyway (at ludicrous MSRPs might i add) in a vacuum its a cool technology, but this isnt an engineering proof-of-concept sort of discussion, so im setting aside that aspect of my opinion as its irrelevant in the face of the lack of functional improvement in my gaming experience (the entire supposed point of the "feature"). i hate motion blur - the frames are blurry and artifacted. i hate input lag - the frames dont respond to peripheral input. what do i gain from using it? nothing.

Its useful for high end cards only! You need base 90fps for fg to feel decent... So yea, its useless for 95% of people

I just set up FSR frame gen for Alan Wake 2 on my 2060 Super an hour ago, works great! Almost 60 fps with medium-high settings, med RT with ray reconstruction and DLSS quality. Without frame gen it's around 35. Oh and no increased latency whatsoever.

Extrapolation is already done for 4-5 years in VR with Asynchronous Timewarp.

The point you are making at 20:15 seems really REALLY bad for the future. Why would you rather dedicate PC hardware towards interpolating and extrapolating frames rather than just rendering the game. At that point you are spending more resources to increase the frame rate number than you are to render the game engine and *play* the game. This is reductive at best. Upscaling and Frame generation are duct tape for poorly optimized games or weaker hardware. Its OK to tape over a few holes, but it becomes a problem when all you are doing is covering everything you see with tape.

I think it sucks. You can't turn 30fps into a convincing 60fps anyway, so this is pointless for the one use case that would have been nice, as many excellent games, mainly older titles, are stuck to 30fps. Sure you can turn 60 FPS into 100FPS and with some luck you won't see some obvious artifacting everywhere and double images on hud elements, big whoop. This whole thing is just yet another ploy to fake higher performance that don't actually exist, but the price increases are absolutely going to be very real. So screw both Nvidia and AMD for this bs.

Minimum base frame rate for FG to make sense is 45 - making 90-100 frames and upwards. 60+ base framerate is optimal and all of this for single player games.

I tried the FSR 3 mod for Cyberpunk in tandem with DLSS on my 3080. The gains are ridiculous, and NVIDIA ought to be ashamed of themselves for being so anti-consumer. If I bought a 40 series card to use frame generation, I'd be pretty miffed right now.

I don't like it, seems like a way for these corps to charge more for less silicon and true performance.

I think FG techniques are awesome, what I won't ever ever agree is on greed and that is what Nvidia did by lying and still lying that DLSS 3 is only capable on 40 series cards, and the only way to enjoy this tech is by buying their over expensive cards while making think that owners of a 30 series cards they have obsolete cards meanwhile that's far from reality, Nvidia didn't think wisely their approach in how to sell their tech, take for example the 4060 and 4060 ti, a 3070 rtx is far superior that both of those cards, so knowing this, Nvidia is telling us a far superior card like the rtx 3070 easely above of the 4060 and 4060 ti is not capable of running DLSS 3 ?, what? all the metrics shown in dozens of videos the 3070 is above those cards and Nvidia is saying is not capable? because "new" generation of tensa cores whatever shat?, hmm yeah if someone still want's to believe that well keep falling for that lol, the FSR 3 mod it literally proves the DLSS 3 only in 40 series card is complete BS!! Also maybe is not on topic but I always wanted to say this about Nvidia greedy ways, do you know why Nvidia raised the prices up so much with the 40 series cards?, well the reason was because crytominers don't have any use of their GPUs now , so what Im trying to say is that Nvidia was always in favor selling literally all their cards to this people non gamers, so the only way possible Nvidia can retain high amounts of income now was to skyrocket their prices, there isn't even a huge jump between the 30 series and 40 series yet again Nvidia catapult their prices without any justification. Very sad Nvidia become this company of pure greediness instead of being a consumer friendly company like it seems it was in the past.

I’ve tried it on cyberpunk, it runs on my 3060 12gb and I’m just amazed. I initially expected noticeable input lag. I was totally wrong, if anything it feels better.

Dumb question.. I have a RX 6800.. how do I enable FSR 3 Frame Generation?