• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

More DLSS...

FYI all new DLSS features trickle down to 40 series, except for the extra framegen frames.

same with 30 and 20 series. They get the enhancements, with no framegen at all.
 
1736223504981.jpeg
 
chameleoneel said:
FYI all new DLSS features trickle down to 40 series, except for the extra framegen framesame with 30 and 20 series. They get the enhancements, with no framegen at all.
Click to expand...
They probably won’t for multi frame. Nvdiia has a habit of excluding last gen from new dlss. They did the same last gen. Just too much money to be made from software alone. How else could he say 5070 = 4090
 
NKD said:
They probably won’t for multi frame. Nvdiia has a habit of excluding last gen from new dlss. They did the same last gen. Just too much money to be made from software alone. How else could he say 5070 = 4090
Click to expand...
That's what I said ;)

"Except for the extra framegen frames"
 
NKD said:
They probably won’t for multi frame. Nvdiia has a habit of excluding last gen from new dlss. They did the same last gen. Just too much money to be made from software alone. How else could he say 5070 = 4090
Click to expand...
The Tensor cores in the older gens literally can't support those new features. "But, Ngreedia!"
 
I said this recently, before seeing the new multiple frame gen in dlss4 marketing:


500Hz or even 1000Hz would be nice in the future some year, but I have no interest in 1080p (or even 1440p) resolution at this point. I think frame gen would have to advance a lot for 4k or greater resolutions for very high Hz screens, to where frame gen would insert multiple frames between each "real" frame accurately. A system where a game engine's moving entities and forces, and your peripheral manipulation + OS drivers/systems, could all broadcast their vectors to a more advanced frame gen tech, informing actual vectors to the tech of all of the vectors in play, instead of solely guessing vectors, where now frame gen is looking at two frames by buffering a frame and comparing the two to guess vectors (which isn't nearly as accurate on it's own).

For example, If I could hit 120fps at 4k or more with "very high+" settings ( to more and more modern ultra settings across years), and have frame gen multiply that by x5 adding 4 frames to reach 600 fpsHz accurately with some more advanced overall game dev, peripheral driver and os dev, + gpu dev frame gen system using informed vectors some-year, that would really be something. 200fpsHz on some future gpu x5 (generating +4 frames) could get 1000fpsHz.
Click to expand...

Link people have been showing already in this thread , timestamped below in the quotes, from the nvidia press releases about DLSS 4. Frame Gen is still not using a game engine and peripheral drivers that inform the system of actual vectors, but they claim that it is so much better than previously that frame gen will be able to produce multiple frames.


View: https://youtu.be/qQn3bsPNTyI?t=81
Click to expand...


"allows to" ... "generate up to three additional frames"

1080120_firefox_ZNsaN5Wo04.png



. . . . .

That means something like 120fpsHz average at 4k could get 480fpsHz average, but they are probably aiming to up the frame rates so people use raytracing more.

A 5080 or 5090, being more powerful, would do better and with dlss4 and framegen upgrades, but in the image below from techspot, a 4090 with RT ultra and DLSS quality is getting 79 fps in cyberpunk. Idk if that means it's using frame gen +1 frame, which would be based on something like 40fps if so.

Personally, my gut feeling is that 100fps (after DLSS upscaling) would be a more optimal foundation of "native" motion articulation for frame gen to work from, delivering 200fpsHz, 300fpsHz or 400fpsHz (average) with the +1 to +3 frames added, for higher Hz screens, but they are going to market it to amplify much lower fps use I'm sure. As I understand it, DLSS and Frame Gen work better the higher the base frame rate and the higher the resolution they are operating on, because there is less degree of difference (change) between action states of two compared frames at higher fps, and at higher resolutions there is more detail to upscale a step down from with dlss, and the pixels are much smaller vs occasional fringe artifacts.



100fps average graph in games might span something like this, with some games tighter than others:

(70)85fpsHz <<<<< 100fpsHz >>>>> 115fpsHz (130fpsHz)

so with framegen +2 it would triple that, +3 would quadruple it.

The presentation in the video shows that there is a drop down in the nvidia panels where you will be able to manually set Frame Gen between +1, +2 and +3 frames. I'm guessing you can use lower if the higher multiples don't operate as cleanly for some reason. (An informed frame generation/amplification system additonally providing actual vector information from peripherals, game world entities and forces, etc. would still operate more cleanly and accurately than AI relying solely on guessing what vectors are in operation between buffered frames).


firefox_7HF1bHBxlk.png



1080122_CP2077_4K_RT.png
Click to expand...
 
Last edited:
mullet said:
I dont think my old eyes could tell.
Click to expand...
lol... I was building PC's in the 90's if that tells you how old I am... and i can 100% tell the difference between 144 FPS and 60 FPS... :ROFLMAO: But to be fair, they did do a medical study awhile back that some people can actually see more "frames" if you will than others, so maybe you are on the low end of that unfortunately.
 
I'm hoping my next screen will be 240hz. I think doubling the Hz is a good jump, halving the sample and hold blur again vs 120fpsHz, and doubling the motion articulation increase again compared to 120fpsHz, ( 4x vs 60fpsHz solid ). That at least when running high enough frame rates on certain games/settings.

I've been following along in a few threads on screens I have some interest in, but a screen upgrade would be after my main rig upgrade (mobo, cpu, ram), which shouldn't be too crazy money wise since I have some applicable parts already. I'll likely get a 5000 series later, perhaps 5090, when price/availability vs rush and scalping is normalized and I'm set up properly for it overall. Without finding a suitable screen upgrade to my tastes in 2024, the timeline for the 5000 series release has put the screen upgrade on the back burner for now in my case essentially so I'm still on a 4k 120hz oled for now.

This might give some insight in regard to higher fpsHz capability. Personal taste is one's own, and eyesight, sensory acuity, etc. but the improvements are measurable in image clarity and motion definition+smoothness.


Sample and hold blur, aka image persistence, affects the entire game world when moving your FoV in 1st/3rd person games mouse-looking, movement-keying, controller panning.
Increasing the fpsHz (feeding a higher Hz monitor enough fps), reduces sample and hold blur considerably. (It is also limited by a screen's response time, and overdrive setting's performance if any.)
At 60fps, the viewport is smearing blur when moving the viewport around at speed.

1079938_blurbusters_sample.and.hold.blur.by.fpsHz_1.png



Pursuit camera recordings give a good representation of how much sample and hold blur you'd see at different fpsHz rates.

This is a pursuit camera of a 60Hz screen at 60fps+ compared to higher fpsHz rates.
Remember that when moving a 1st/3rd person viewport around at speed, the entire game world full of high poly objects, high detail textures and depth via bump mapping, architectures and geography, fx, etc will be blurred, not just a cell shaded cartoon like ufo icon.

240hz-motion-blur-jpg.301316




This is a pursuit camera image of a 480Hz OLED at 480fps, by Mark R of blurbusters.com.

480hz-oled-pursuit-camera-clearest-sample-and-hold-oled-ever-v0-wf9s5c8uopcc1.png





Higher fpsHz also provide more motion pathing articulation and even increased animation state stepping, like an animation flip book filled with more unique animation states, with the book flipping faster.

blurbusters_mouse.stepping.by.Hz_2_a.jpg


blurbusters_mouse.stepping.360Hz_1.jpg




high.fps.hz_running.cartoon.animation-2.rev..jpg
 
Last edited:
III_Slyflyer_III said:
lol... I was building PC's in the 90's if that tells you how old I am... and i can 100% tell the difference between 144 FPS and 60 FPS... :ROFLMAO: But to be fair, they did do a medical study awhile back that some people can actually see more "frames" if you will than others, so maybe you are on the low end of that unfortunately.
Click to expand...
We are probably the same age, bumping 60. I guess I need to upgrade my old dell ultrasharp ips 27" (mainly for PS).

EDIT: new monitor coming we shall see. only 165hz.
 
Last edited:
Welp looks like DLSS is the path forward.. all hail our AI overlords lol. Pretty soon the AI will be making the game up on the fly with DLSS.. I mean it's doing that now but more in a procedural generation way.. idk.
 
revenant said:
Welp looks like DLSS is the path forward.. all hail our AI overlords lol. Pretty soon the AI will be making the game up on the fly with DLSS.. I mean it's doing that now but more in a procedural generation way.. idk.
Click to expand...
Just wait till AI games interface with chips in our brains... we will not even know what's real anymore!
 
kac77 said:
What I love is that in an effort to sell FG in DLSS4, nVidia is now showing all of the artifacting of DLSS 3 that never gets shown anywhere (even in reviews). I'm assuming because it's "better than native".:rolleyes:
View attachment 702361
Click to expand...

What? Alex from DigitalFoundry has shown the ghosting problems not just with FG but SR as well plenty of times ever since v2.0.
 
kac77 said:
What I love is that in an effort to sell FG in DLSS4, nVidia is now showing all of the artifacting of DLSS 3 that never gets shown anywhere (even in reviews). I'm assuming because it's "better than native".:rolleyes:
View attachment 702361
Click to expand...

I've seen DF (publisher of that vid you quoted) point out DLSS shortcomings and artifacts countless times across many game/hardware reviews and tech analysis.

I don't play a ton of games, but the few games I've played that support DLSS, esp if the game doesn't have any AA options like Doom Eternal and System Shock that I'm playing currently, DLSS quality actually improves image quality in the areas I noticed most. Like horizontal grating on floors that are often aliased and warp in a distracting manner.

I'm still on a 12GB 3080 though, so haven't messed with any frame gen stuff yet, so interested to see how that actually looks and feels in terms of latency. From what I've gathered, I wouldn't expect any real noticeable latency on any game that has a base frame rate of 60 at least.

But now I can't wait for DLSS 5 or 6 when it can generate entire games to play from a single frame. 😁
 
T4rd said:
I've seen DF (publisher of that vid you quoted) point out DLSS shortcomings and artifacts countless times across many game/hardware reviews and tech analysis.

I don't play a ton of games, but the few games I've played that support DLSS, esp if the game doesn't have any AA options like Doom Eternal and System Shock that I'm playing currently, DLSS quality actually improves image quality in the areas I noticed most. Like horizontal grating on floors that are often aliased and warp in a distracting manner.

I'm still on a 12GB 3080 though, so haven't messed with any frame gen stuff yet, so interested to see how that actually looks and feels in terms of latency. From what I've gathered, I wouldn't expect any real noticeable latency on any game that has a base frame rate of 60 at least.

But now I can't wait for DLSS 5 or 6 when it can generate entire games to play from a single frame. 😁
Click to expand...
You're correct on frame generation. I think a lot of people are discounting the effectiveness of dlss and its anti aliasing effects too.
 
T4rd said:
I've seen DF (publisher of that vid you quoted) point out DLSS shortcomings and artifacts countless times across many game/hardware reviews and tech analysis.

I don't play a ton of games, but the few games I've played that support DLSS, esp if the game doesn't have any AA options like Doom Eternal and System Shock that I'm playing currently, DLSS quality actually improves image quality in the areas I noticed most. Like horizontal grating on floors that are often aliased and warp in a distracting manner.

I'm still on a 12GB 3080 though, so haven't messed with any frame gen stuff yet, so interested to see how that actually looks and feels in terms of latency. From what I've gathered, I wouldn't expect any real noticeable latency on any game that has a base frame rate of 60 at least.

But now I can't wait for DLSS 5 or 6 when it can generate entire games to play from a single frame. 😁
Click to expand...
Doom Eternal uses TAA and it can't be turned off in the game. Turning on DLSS supersedes TAA. DLSS is still a form of TAA, but it looks better because it does a better job at approximating the final color information in the reconstructed frame.
GoldenTiger said:
You're correct on frame generation. I think a lot of people are discounting the effectiveness of dlss and its anti aliasing effects too.
Click to expand...
It is certainly preferable to standard TAA.
 
As I understand it, DLSS and Frame Gen work better the higher the base frame rate and the higher the resolution they are operating on. (The old saying "you can't get blood from a stone" ?). The reasoning given being because there is less degree of difference (change) between action states of two compared frames at higher native fps, and at higher resolutions there is more detail to upscale a step down from with dlss, and the pixels are much smaller vs occasional artifacts.

If that's true, then conversely, the lower your base frame rate then the higher the % change will be between two frames ( ~> the bigger the leap to guess the manufactured frame?). Perhaps also, the lower the frame rate, the longer an imperfect guess frame would be visible on the screen before the next "real" frame is shown.

Also, if a given, the lower the base rez upscaled from and the lower the final rez (e.g. 1080 -> 1440 rather than 1440->4k), the less detail DLSS has to upscale from. The reasoning also follows that the larger the pixel granularity, then the more obvious things like artifacts and fringing might be (similar to how text fringing is much more obvious at lower PPD vs higher PPD).

Point being that there may be some variables which could make blanket staments -about how much artifacting might be visible- not accurate to every configuration. . pc hardware, and between different types of games, selected settings/performance/quality, and different game titles.

. . .

I've also seen articles showing how DLSS+FrameGen tech can get confused by orbiting 3rd person cameras, because it might look like the character isn't moving at times when both the character and the camera are moving independent of each other, confusing the ability of the AI to guess the vectors in play accurately. That's another reason why a system which informed actual vectors(game engine broadcasting vectors of in game entities, forces, in game virtual camera, etc., and player peripherals broadcasting player input vectors,) could be better since it would inform the system of actual vectors in addition to comparing buffered frame(s).
 
Last edited:
elvn said:
As I understand it, DLSS and Frame Gen work better the higher the base frame rate and the higher the resolution they are operating on. (The old saying "you can't get blood from a stone" ?). The reasoning given being because there is less degree of difference (change) between action states of two compared frames at higher native fps, and at higher resolutions there is more detail to upscale a step down from with dlss, and the pixels are much smaller vs occasional artifacts.

If that's true, then conversely, the lower your base frame rate then the higher the % change will be between two frames ( ~> the bigger the leap to guess the manufactured frame?). Perhaps also, the lower the frame rate, the longer an imperfect guess frame would be visible on the screen before the next "real" frame is shown.

Also, if a given, the lower the base rez upscaled from and the lower the final rez (e.g. 1080 -> 1440 rather than 1440->4k), the less detail DLSS has to upscale from. The reasoning also follows that the larger the pixel granularity, then the more obvious things like artifacts and fringing might be (similar to how text fringing is much more obvious at lower PPD vs higher PPD).

Point being that there may be some variables which could make blanket staments -about how much artifacting might be visible- not accurate to every configuration. . pc hardware, and between different types of games, selected settings/performance/quality, and different game titles.

. . .

I've also seen articles showing how DLSS+FrameGen tech can get confused by orbiting 3rd person cameras, because it might look like the character isn't moving at times when both the character and the camera are moving independent of each other, confusing the ability of the AI to guess the vectors in play accurately. That's another reason why a system which informed actual vectors(game engine broadcasting vectors of in game entities, forces, in game virtual camera, etc., and player peripherals broadcasting player input vectors,) could be better since it would inform the system of actual vectors in addition to comparing buffered frame(s).
Click to expand...
It isn't necessarily that frame gen works "better" at a higher base framerate, but that the input latency is based on the framerate you were getting before frame generation. That is why the Cyberpunk video was still showing 35ms at 200+ FPS because that is how long a frame lasts at 28 FPS. It's similar to how you explain the resolution that is being upscaled from.
 
I get that about latency, very interesting for sure and makes total sense. Very neat.

What I was questioning, and suggesting, though is:

========================================================

.

..whether the resulting quasi-frame generated is more % accurate (vs artifacts/inaccuracies) when there is less % difference between frames to guess between. That due to higher frame rate cutting up the world state action into finer slices with less difference between each to change.



.. That and, whether the % inaccuracy of quasi/tween frames might be less apparent to the viewer when displayed for a much shorter period between a considerably higher base frame rate's frames as compared to a lower base frame rate displaying the % inaccurate frames for longer.



..also, whether higher resolutions would make the visibility of some of such % inaccurate generated frame's inaccuracies less apparent by some % due to higher PPD (like text fringing at lower PPD vs higher PPD)


. . . . .

..also, for DLSS upscaling, whether it has less % inaccuracy when upscaling from higher resolutions which have more detail to start with (e.g. up from 1440p rather than 1080p), and that as it works hand in hand with frame generation for an overall % inaccuracy between "real" upscaled frames and "quasi/tween" generated frames.

. . . .

edit:

.. and , as brought to attention, - that the input lag will be lower when running a higher base frame rate since the input lag remains based on that rate rather than the resulting frame gen amplified one
 
Last edited:
mullet said:
I can't tell any difference with anything over 60fps.
Click to expand...
Main stream visibility need geometrics and GtG=0.

Such as 4x differences (e.g. 120 vs 480) and GtG=0 (e.g. OLED)

You gotta VHS-vs-8K it temporally, not the mere equivalent of 720p-vs-1080p much like 120Hz-vs-144Hz, or 240Hz-vs-360Hz LCD (1.5x diluted to 1.1x due to nonzero GtG).

Pro users and esports can benefit from small Hz differences, but mainstream / office needs large geometrics for a good upgradefeel. The blissfully unaware mainstream that say they can't tell apart 720p versus 1080p, definitely could tell apart 120fps vs 480fps on an OLED during a 2000 pixels/sec panning framerate=Hz test.

Also the world's most common 120Hz display are generally the overdriveless Apple LCD's, where 60fps vs 120fps is marginally visible unlike on an OLED. Once you go OLED, 60fps vs 120fps becomes much more dramatic, and 60fps vs 240fps even more so, even to the mainstream.
 
elvn said:
As I understand it, DLSS and Frame Gen work better the higher the base frame rate and the higher the resolution they are operating on. (The old saying "you can't get blood from a stone" ?). The reasoning given being because there is less degree of difference (change) between action states of two compared frames at higher native fps, and at higher resolutions there is more detail to upscale a step down from with dlss, and the pixels are much smaller vs occasional artifacts.
Click to expand...
Yes, I recommend base framerates above flicker fusion threshold, depending on how fast your display is.

That means pre-framegen framerates of at least ~60fps on LCD, and at least ~80fps on OLED.

This forces oscillating artifacts (framegen-vs-nonframegen) to be hidden by its own blur (flicker fusion)

This dramatically reduces the visibility of framegen artifacts.

The Developer Best Practices section of my famous 10:1 framegen article (that correctly predicted many elements of DLSS 4), also talks about having high base pre-framegen framerates.

kac77 said:
What I love is that in an effort to sell FG in DLSS4, nVidia is now showing all of the artifacting of DLSS 3 that never gets shown anywhere (even in reviews). I'm assuming because it's "better than native".:rolleyes:
Click to expand...
I was at CES 2025 and was part of the media scrum that saw the DLSS demos. I'm biding my time creating my piece obviously -- I don't rush pieces out like the mainstream media, (and Blur Busters is nowadays more of a test-creation lab rather than a media writer anyway).

However, some short tidbits:

What I saw of DLSS 4 is that the fake frames looked better than the fake frames of DLSS 3 and much fewer artifacts than TAA.

If we must be picking our framerate poisons for framerate-based motion blur reduction (instead of flicker-based motion blur reduction), then improved framegen is one of the least-pick-poison methods of getting your framerate's worth out of 4K 240fps RTX ON graphics on today's 4K 240Hz OLEDs.

More of the visuals *did* get preserved than during DLSS 3.5:
- The fake frames looked better; *and*
- The large ratios involved (4x+) often meant the motion clarity improvement (especially on OLED) started to significantly exceed any minor-to-almost-unnoticeable spatial clarity loss
- This was much more apparent on OLEDs than LCDs, as OLEDs benefit far more from framerate-based motion blur busting than LCDs do;

It blew totally past UglyTAA in motion quality, that is for sure.

It's not perfect. There were some artifacts. But it is getting down to roughly about 0.1% of year 2010-era Sony Motionflow Television Interpolation.

Whether that's good enough or not, is for you to decide; that's why I also released the CRT simulator for my strobe-based blur reduction fans.

I play both the strobe-based blur busting and framerate-based blur busting games.

Here's the Law of Physics, from Blur Busters Law:
Strobed = motion blur is pulsewidth (assumes squarewave, framerate=Hz)
Unstrobed = motion blur is frametime (assumes GtG=0)

Therefore, we need large framerate increases (4x-8x) for framegen to gigantically reduce display motion blur more than strobing. You need 480fps 480Hz on an OLED (2ms MPRT) to have sample and hold with less motion blur than early strobe backlights such as LightBoost (2.4ms MPRT).

Longtimers remember the younger me extolling about LightBoost (oh boy...)

Yes, that means I've seen RTX ON graphics on sample and hold with less motion blur than a LightBoost strobe backlight at (LightBoost=100%). It's impressive how far framerate-based blur busting has come; and starting to reach the early echelons of strobe backlights in motion quality -- while staying Ergonomic FlickerFree. Yes, we're having to pick poisons to do so, but TAA needs to die in a flaming fire, while framegen needs to improve even further, add another zero, from today's 0.1% artifacts, to tomorrow's 0.01% artifacts of 2010-era interpolation.
 
Last edited:
I got the new monitor lol not OLED)
Chief Blur Buster said:
Main stream visibility need geometrics and GtG=0.

Such as 4x differences (e.g. 120 vs 480) and GtG=0 (e.g. OLED)

You gotta VHS-vs-8K it temporally, not the mere equivalent of 720p-vs-1080p much like 120Hz-vs-144Hz, or 240Hz-vs-360Hz LCD (1.5x diluted to 1.1x due to nonzero GtG).

Pro users and esports can benefit from small Hz differences, but mainstream / office needs large geometrics for a good upgradefeel. The blissfully unaware mainstream that say they can't tell apart 720p versus 1080p, definitely could tell apart 120fps vs 480fps on an OLED during a 2000 pixels/sec panning framerate=Hz test.

Also the world's most common 120Hz display are generally the overdriveless Apple LCD's, where 60fps vs 120fps is marginally visible unlike on an OLED. Once you go OLED, 60fps vs 120fps becomes much more dramatic, and 60fps vs 240fps even more so, even to the mainstream.
Click to expand...
I do have a LG OLED 77" and yeah I can tell the difference between 720, 1080 and 4K. The 4K is the most obvious.
The biggest to me is when I record video at 30 vs 60 in panning left and right. Drives me nuts when people make videos @ 30 and pan fast around makes it look like a slide show.
Yeah I understand the math in what you are saying. Yes if you are competitively playing CS then yes the Hz are KEY. The higher the Hz (if the FPS are matched or over) the more frames and smoothness makes it easier to hit the target in motion.
I just got a new monitor this week with 165Hz and the immediate thing I noticed is scrolling my mouse on web pages butter smooth. Playing games I noticed with no vsync, uncapped fps, freesync on, it is still not as smooth to me as say 60 or 75 vsync on.
 
mullet said:
I do have a LG OLED 77" and yeah I can tell the difference between 720, 1080 and 4K. The 4K is the most obvious.
Click to expand...

I mean...on a 77" display, no shit. Most people are gaming on displays between 27 and 32 inches, though, with the same resolution(s). Lot harder to tell the difference there.
 
MavericK said:
I mean...on a 77" display, no shit. Most people are gaming on displays between 27 and 32 inches, though, with the same resolution(s). Lot harder to tell the difference there.
Click to expand...
Depend on the sitting distance and eyes, face on the monitor, same game running at the same time on a 1440p and 4k, with 27 inches I was not able to tell the difference, same for really high quality video, while text is obvious
 
BlurBuster... I'm kind of wondering if when we reach 500hz OLED being common is there a combination of shooting for say 120hz locked from the game, and using your CRT scan shader at 480hz to kill blur? Will that tech still work with base frames above 60hz for a best-of-all-worlds approach? Or some other ideal combination?
 
LukeTbk said:
Depend on the sitting distance and eyes, face on the monitor, same game running at the same time on a 1440p and 4k, with 27 inches I was not able to tell the difference, same for really high quality video, while text is obvious
Click to expand...
Correct me if I'm wrong, but doesn't cloning in windows just scale the higher resolution up for the second monitor if you're doing that?

I've had a bunch of 4k and 1440p monitors, and in back to back gaming tests the resolution difference was obvious, but when cloning it appeared the same.

This was across 24" 4k, 27" 4k and 1440p, 28" 4k, and 32" 4k where I could see the pixels in games.

How were you testing the same game on both screens at native resolutions exactly?
 
You must log in or register to reply here.
Back
Top