Why do people not like Frame Gen and other things to do with DLSS and such?

[Cinlow]

Call me crazy, but I like Nvidia's DLSS, Frame Gen, AI AI-powered tools. It adds ways to play games at higher frame rates to match monitors, allowing older cards like the 30 and 40 series to last longer than they should, and even making the card with less VRAM slightly worth it.
It highlights some of the potential with AI as well. AI hasn't been explored much, but DLSS and Frame Gen show how AI could be used positively.
I think the only argument that could be made is that the cards are too expensive, which is true.
I am not saying "Oh, Nvidia on top!" I am saying, "I like how they positively use AI."
I dislike that Nvidia is leaning more on AI and not gamers anymore, and it is a sad reality.
I dislike that people talk trash on Nvidia for choosing to use AI to power their cards.

[WereCat]

4 minutes ago, Cinlow said:

Call me crazy, but I like Nvidia's DLSS, Frame Gen, AI AI-powered tools. It adds ways to play games at higher frame rates to match monitors, allowing older cards like the 30 and 40 series to last longer than they should, and even making the card with less VRAM slightly worth it.

thats what they in theory are supposed to do but in practice dont

 

they use more vram

frame gen becomes most useful when you already have plenty of performance and has major drawbacks on lower framerates… if that doesn’t bother you then more power to you 

[ewitte]

Framegen makes the fps faster but the response slower.  The lower the base framerate the worse the experience.  Try running 4x FG on a game getting 30fps it would feel like you're walking through quicksand.  It is not passible until your base framerate is over 60-90fps depending on your sensitivity.  The slower cards are likely to get the least measurable improvement.

[Vince492]

I like framegen, in VR, it is the only way to get a smooth experience and nice graphics in games like DCS, X-Plane 12, and msfs202x. I use the quest 3 AWS and its fine, sometimes there is artifact, like the helicopter`s rotor making waves, but after a few minutes I don`t see it anymore.

 

I tried ASW, in racing games like ACC and it is mostly fine. Works better if you do it at 120hz and you can maintain 60fps, I'm not good enough to see a latency when doing 36fps with ASW on, but I prefer having real 72fps rather than 60 real one and 60 fake ones.

[GuiltySpark_]

Primarily it stems from and then gets regurgitate by people with a combination of no personal experience and/or using the tech outside of optimal and recommended conditions. 

 

Examples.  1080p users trying to use DLSS Performance and complaining about the image quality. Or 4070 users who get 45fps native trying to use Frame Gen to improve their framerate and complaining about the input latency penalty. 

 

Then there are the people that just live in the past and hate on new things to hate on them.

 

I've been a user and believer of this tech since it was released and will continue to make use of it as its massively improves my gaming experience when used properly.

 

 

 

 

[porina]

Before RTX era, pretty much everything was directly rendered. Some people attach some kind of purity to that, and see upscaling and framegen as tricks.

 

But if we take that historical era as reference, personally I do find great value in upscaling. I tend to only use it at the higher quality settings, where it is often practically not noticeable, while giving a nice boost to frame rates. At lower settings the impact is more visible.

 

Frame gen is a bit more complicated. It can give better latency if you compare native rendering without modern latency reduction methods, but worse than native with that latency reduction. Still, if old native rendering was acceptable, frame gen latency can be too. To me latency is pretty much a good enough, or not good enough thing. If it feels good, I don't care if it is 1ms or 100ms. If it is bad, I don't care if it is 100ms or 1000ms.

 

I first tried Nvidia frame gen with Portal RTX. I had an under-powered 3070. FG did give a fps boost, but it felt a bit weird. I didn't have quite enough base fps to keep latency in the comfort zone. When AMD joined in with FSR3 FG, I tried that in Forspoken demo. That felt ok in responsiveness, but unfortunately the mandatory FSR upscaling looked awful. More recently, I tried HL2 RTX demo on 5070 Ti. Here, I could get an acceptable base frame rate with upscaling only, and in this scenario FG did help with the fluidity. Latency was well in my "good" zone, and turning on FG gave me that higher fps smoothness. It wasn't perfect though, with artefacts showing on UI elements when FG was turned on as the interpolators can't separate the UI from world content. I'm not sure if it can be fixed in this instance, since the preferred way to do it is to draw the UI separately after processing the world content. If HL2 was never intended to work that way, it might not be possible. This should not be a problem for any game that natively supports FG.

[Hinjima]

Even though I have a high end GPU ( RTX 4080 ) I still use DLSS Quality whenever I can.

I really do enjoy DLSS and it helps improve framerate without really affecting the visual quality in my opinion.

 

[Jurrunio]

Imo all that is caused by Nvidia's disgusting and out of touch marketing. Yes frame gen and DLSS improve frame rate more while sacrificing quality less than the old way of turning down settings, but they are not equivalent to native rendering. Putting new cards with new tech turned on in a direct fps comparison to older cards is fine, but that's not showing the whole picture so setting the marketing language and tone 100% according to those comparisons is not fine if not outright scummy. "RTX 5070, 4090 performance" is brainrot speech.

 

Imo all they really have to do is to show what settings are needed to make GPUs in the comparison get to the same frame rate at the same resolution, then show people the differences in graphical quality on the receiving end. This is the value of FG and DLSS, not just fps numbers and especially not fps to price since price is as unstable as it gets right now.

[Mark Kaine]

What ai? Lol...

 

DLSS... Looks trash most of the time...framegen is more useful but has latency issues and sometimes looks like trash too...

What's not to like, right... ?

 

Or maybe they shoulda put all that efforts into raster performance instead of "Raytracing" DLSS, etc etc... then devs could focus on making good games instead of shoveling the newest "features" (which they neither understand or care about) into everyone's faces...

 

I always find it weird when people say observations or even indifference are "hate"... Makes me wonder what their agenda really is... but sure, good guy Nvidia, I guess??

 

 

3 hours ago, porina said:

Before RTX era, pretty much everything was directly rendered. Some people attach some kind of purity to that, and see upscaling and framegen as tricks.

Well, they are... crutches to get the way too demanding "Raytracing" feature to work at "acceptable" performance... Remember they said their "RTX" shenanigans would run on dedicated "RTX" cores and basically cost no performance??

 

(I don't either really but that's what their marketing basically implied)

 

It's really all mostly terrible marketing with buzzwords - but what really gets me is that people don't seem to understand that while some effects might use Raytracing, that's not really how Raytracing should be used as you can do that easily with other, less demanding "tricks"... Raytracing should be used as rendering pipeline getting rid of any and all pixels, and aliasing... Ofc, we're probably still a decade or two away from cards being powerful enough for that.

 

As said it's not really that fact alone, it's mostly the marketing, the bad optimization because of these features and devs obviously abusing things like DLSS and framegen as substitutes for basic game optimization.

 

 

[SorryBella]

5 hours ago, Cinlow said:

and even making the card with less VRAM slightly worth it.

As a hardline backer of Frame Generation (currently using LSFG 3.0) and temporal generation (pingponging between LS1 and XeSS Quality), no. VRAM is definitely where its a hard case of "There's no replacement for displacement" for GPU, especially in the new paradigm where textures are directly loaded to VRAM and computer RAM for faster load times and smoother performance. Lower than 12GB (hell some titles like Black Myth Wukong and MH: Wilds easily exceeds that into 16GB realm on 4K) is already insufficient, and game with this in mind quickly suffers in 1% and .1% low (read: more stutters) because more swaps to the SSD have to be done.

 

And as everyone points out, the latency issue is still terrible for a lot of people and even the best efforts are still +15ms from native. For tightly timed genres like rhythm games, bullet hells, and souls like, its unacceptable. And another thing, Frame Generation also dont cover you when you are heavily GPU bound, if youre at 100% and then you turn on frame generation its a stutter fest, and it shows not only in feeling but also in graphs if you run CapFrameX to find out.

[starsmine]

3 hours ago, Jurrunio said:

Yes frame gen and DLSS improve frame rate more while sacrificing quality less than the old way of turning down settings, but they are not equivalent to native rendering. 

I want to stress that in almost every situation where DLSS looks "better" then native.

The cause is actually bad TAA implantations at native. 

[DeerDK]

8 hours ago, Cinlow said:

Call me crazy, but I like Nvidia's DLSS, Frame Gen, AI AI-powered tools. It adds ways to play games at higher frame rates to match monitors, allowing older cards like the 30 and 40 series to last longer than they should, and even making the card with less VRAM slightly worth it.
It highlights some of the potential with AI as well. AI hasn't been explored much, but DLSS and Frame Gen show how AI could be used positively.
I think the only argument that could be made is that the cards are too expensive, which is true.
I am not saying "Oh, Nvidia on top!" I am saying, "I like how they positively use AI."
I dislike that Nvidia is leaning more on AI and not gamers anymore, and it is a sad reality.
I dislike that people talk trash on Nvidia for choosing to use AI to power their cards.

My first encounter with DLSS was when I tried turning on Raytracing in Cyberpunk. Suddenly every texture had this "oily" look to it, and I then discovered that DLSS had also been activated without my approval. It wasn't a good first impression, and since then I've been reluctant to turn it on, preferring native resolutions with a clearer picture to upscaling, and balancing other features to preserve an acceptable fps. 

 

Haven't really played around with framegen as I have not been in a situation where I need it.

 

The thing you should probably keep in mind when saying that " the only argument that could be made is that the cards are too expensive" is also the backlash they caused when they twisted the truth and claimed that the 5070 would deliver 4090 performance and such. And not really going Mea Culpa when called out on these claims, but rather doubling down.

AI generated frames are not the same as natively rendered. 

While the AI (I really do hate the widespread use of that word) development is very interesting, and holds potential, cramming it into every product and go "there, its better now because its with AI" may not be way to go.

[porina]

Reading the comments in this thread, some are focusing on worst case scenarios. These are not a universal solution, so there will be cases where it is not appropriate to use. What if I told you, you don't have to use it in those situations! However there are solid tangible gains where it is appropriate. Unless you're struggling to run 1080p in the first place, non-AMD upscaling generally is mature. Frame gen is still maturing so that is more situational.

[DreamCat04]

To be honest, I also fell onto a bit of a hate-train since all I experienced was the high input latency mess from using AMD's FMF on my ROG Ally where I only ever got a base framerate of between 40 to 60 FPS where the input latency added by AMFMF was really noticeable. I stopped using it after a while because I had to re-patch the custom driver every time there was a driver update and I got tired of that lol. I then tried it again a few weeks ago in Cyberpunk on my RTX 4070Ti where with everything cranked, RT & PT on, my base framerate was around 60. It was geniuinely cool having over 120 FPS at times with everything cranked and the very demanding Path Tracing turned on, so I finally got off the hate-train. It's kinda misleading how Nvidia said that the 5070 was gonna have 4090 performance, that's definitely wrong as the 4090 used regular frame gen when the 5070 was using MFG

[ewitte]

13 hours ago, Mark Kaine said:

What ai? Lol...

 

DLSS... Looks trash most of the time...framegen is more useful but has latency issues and sometimes looks like trash too...

What's not to like, right... ?

 

 

I'm not sure what you are looking at.  9/10 times I have to pause the game and pull out a magnifying glass to tell the difference.  Usually, the more obvious differences DLSS is the one that looks BETTER due to not having lost textures or blurry images in TAA.

[xg32]

18 hours ago, Cinlow said:

Call me crazy, but I like Nvidia's DLSS, Frame Gen, AI AI-powered tools. It adds ways to play games at higher frame rates to match monitors, allowing older cards like the 30 and 40 series to last longer than they should, and even making the card with less VRAM slightly worth it.
It highlights some of the potential with AI as well. AI hasn't been explored much, but DLSS and Frame Gen show how AI could be used positively.
I think the only argument that could be made is that the cards are too expensive, which is true.
I am not saying "Oh, Nvidia on top!" I am saying, "I like how they positively use AI."
I dislike that Nvidia is leaning more on AI and not gamers anymore, and it is a sad reality.
I dislike that people talk trash on Nvidia for choosing to use AI to power their cards.

i consider dlss quality to be about the same as native. The problem i have is with frame gen, it doesn't feel the same and the overhead is huge. So i can easily end up in a situation where it is 4k 90fps vs 4k 60->120 FG, and 4k 90 fps feels better overall. 

 

The main usage for now might be for 120fps->240 FG, it's kinda niche.

 

The main problem i have with nvidia is forcing 12vhpwr down our throats and the 50-seires drivers are actually a dumpsterfire.