DLSS vs FSR Input Latency - Which Does it Faster?

cant you test fsr on a amd gpu to compare to this to see if it has lower latincy?

If I keep the FPS on 60(my monitor frame rate), while using DLSS, it will give me more latency? When I disable it, I feel like it has a better response

I have a RTX 2060S

I thought they both add input lag, due to AI need to waste some time to enforce the image. I'm glad I was wrong

Bre no shooter conparison? Gg

What are you talking about? You don't have an input lag about 20ms with 200fps. Wtf?

Dlss on with vsync 60 fps VS Dlss off with vysnc 60 fps, which has less latency?

perfect video !

FSR is already amazing in terms of latency compared to DLSS. It will win the upscaling race for sure because its open source. THanks AMD!

Man I usually play at roughly 16ms of input lag on a quite old PC... how can you achieve such abysmally bad performances on your rig?

Have newer games changed in such a worsening way?

It seems like performance and latency is better with fsr. But aesthetically, which puts out a better image dlss or fsr?

If Thor can fly, why does he need to twirl his hammer and fling it forward to fly forward? Can he only float? Does meowmeow propel him forward.

Waiting on a Linux Proton vs Windows comparison or even just a Linux gaming video to cover what performance might be like on the steam deck when it releases

FSR is free for both, if FSR is equal or better then fuck DLSS. 😊

What amazes me is how little DLSS coverage there was from HUB prior to FSR’s launch. Now, there are so many vids studying the two when really it just comes down to spatial vs temporal methods for image enhancement and upscaling from a lower resolution reference frame(s). At the end of the day, regardless of the quality/performance title… generally equal quality output settings render at roughly the same performance increase… which was the whole point of DLSS (2.0+) implementation but sadly was kept out of any major discussion and wasn’t IMO given enough credit historically when it comes to AMD vs NV in GPU comparisons, especially when you add in RT. It wasn’t until FSR came along that they seemed interested in the technologies.

Very interested in seeing some older games with new hardware... any chances of revisits?

My GTX 1650s like AMD FSR

I have found FSR to be extremely useful for getting VR games to be playable on older or slower hardware, FSR can easily be added to any VR game that uses OpenVR, so most of them, vastly improving the fps and frame time consistency. Performance is extremely important with VR games because inconsistent frame rates cause nausea, so usually the reduction in visual quality is worth the performance increase. DLSS it's not as useful for this because it is not supported in many VR games, cannot be added, and only works with a few GPUs.

I'm using a 3080 and used Nvidia cards with all my builds but I want FSR and AMD to lead a bit more. betting on the underdog is thrilling.

You forgot one thing - Dlss working only on Nvidia cards and only RTX series - AMD working with both card from Nvidia and AMD also working with an old card so FSR win here

fsr looks worse ( higher fps but looks worse ) compared to the same quality settings as nvidia dlss @ quality hence the higher fps but worse image quality

All we need is large scale adoption of FSR in to most of current and upcoming games. It will benefit both AMD and nvidia users.

Lovely! FSR 1.0 can be a replacement to DLSS 2.0 . AMD is born mature 💜

This testing is surprisingly useful to me.
Since I game at 1200p and am not gpu bound, I'll not use upscaling.
Another useful conclusion is, that I can buy either amd or nvidia, while deciding on factors like linux driver support and power efficiency.
And not having to bother with the differences in some proprietary features.

Ooh, Tim, that switch-up in the intro was brilliant! You started the video off with a question and then rolled the logo animation. I like it! It's better!

still not sciencing properly.
first up should have used frametimes instead of framerate. ie native frametime for avengers was 7.46ms, so it's taking just shy of 8 frames to respond to input...wtf is it doing? does it need 8 frames to composite the scene? probably not, there's inherent latency in the hardware being tested so it's probably wasting a bit of time there....so first two unscientific methods not establishing the baseline of the hardware itself and using misleading equivalences, FPS is not a unit of time.

third as per my previous comment in the first video, you didn't test a non sponsored game. whats the control?

fourth, no AMD hardware was tested. just because we have no comparison of amd running DLSS doesn't mean we have no comparison to AMD and GTX hardware running FSR or native input latency. it's possible nvidia sponsored games have twice as much latency OR half as much latency as fidelityFX or unsponsored titles. or that older nvidia hardware is being punished in latency. hell maybe even new hardware at different tiers offer different input latency. comparing dlss to fsr in an nvidia sponsored title is like running a religious debate in a church, that sponsor might have an agenda. we all know that nvidia wont show up to the debate when conditions aren't set by them, nvidia has no interest in making dlss for AMD hardware but you can still test their hardware outside their ecosystem and compare the differences to inside that ecosystem.

which brings to lastly, no controlled software. this seems difficult to do. but unreal engine exists, anyone can build a game using unreal engine and plenty of content creators make rtx games for fun. i'm sure you could collaborate with one of those channels to make the same game 3 times. not the same game with 3 options but 3 separate executables one that's always RTX on/off, one that's always FSR on/off and one that's always running at one of the two target resolutions and has neither system anywhere near it. all that game would need to do is flash the screen when the mouse was clicked. or you could load it up with a bunch of your logo's and have it do something when the mouse was clicked.

it seems a waste to invest the time you did in what amounts to nvidia marketing.

Best Hardware lovers reviews by faaaar !!!

Science!

Dlss must take slightly longer by needing to use a tensor core, not enough to mean anything other than that the higher fps the lower latency no matter the upscaling.

this is the result i expected. the difference is in margin of error territory

+

It's nice to have some hard data on this. I was a little worried about it - but the results here are pretty clear.
It's a relief that we don't have another layer of complication to worry about :)

I do not understand why you did not test the input lag in 4k. In 4k input lag is much more important compared to lower resolutions which most of the time execelnt input lags.

meh? waiting for 120 w max gpu benchmarks. fsr wins btw, its free

I don't play latency sensitive games but your content are always interesting. Great work as always.

Sometimes, in your focus on engineering and data, you forget the consumer perspective, and that is, that FSR is available for AMD as well as Nvidia GPUs.

Anyone notice Rocket League playing so much better on Nvidia? I really wish you guys would do some benchmarks with RL. Always teasing us with footage of the game but it's never included apart from.. 1?! Benchmark. Just saying :)

Hardware Unboxed presents "Moore's law is a Clown: Part Deux"

So I have a question: so you have proven that there is a net gain when using these technologies but I kind of noticed something in the numbers for the initial dlss video so I ask, is the uplift in framerate equivalent to the decrease in latency, or does latency not get as much of a decrease due to the overhead? I get it, the benefits outweigh the detractors, I'm more just academically curious, but also, it would be a good way to measure the overhead of each comparatively to each other.

without watching the video I'd guess not really, and mostly depends on fps

Well input latency is a matter of concern exclusively for competitive titles only, so this video is bulshit

I will say this, the fact that FSR works on practically every wine/proton game in Linux is great. I'm using it on nvidia and it still works flawlessly.

I think it will be of huge benefit when trying to run big AAA titles on the Steam Deck. You can drop the games to 400p, and probably notice very little difference. Especially on such a small screen.

I am testing it at 720p on a 1080p tv. I think 44 inches with a gtx1060 max q laptop. A lot of text looks over sharpened, but the games look like I expect at 1080p. Actually better, because I am able to run higher settings.
Witcher 3 is the best example. Using FSR I can run the game maxed out except hairworks at lock 60, on a max q 1060. I needed medium at 1080p.

Bits make fast computer software make graphics better part as fast as well

8.1 gbps ×2

Input latency is clearly observable in high precision rythym games the best score is between 16.66ms and 21ms in IIDX or ez2on. Adding an additional 20-40ms can make you perform much worse.

Hahaha "This video can be used for when a fight breaks out on forums" that comment was awesome.

I came for latency details and stayed for the most efficient and correct review of "Marvel's Avengers" possible.

Conclusion: No technology can possibly save Marvel's Avengers.

It's growing.. Soon enough I will know you again

6800xt vs 3080 input latency comparison