NVIDIA DLSS 4.5 Dynamic Multi-Frame Generation and 6x Mode Officially Arrive

DLSS 5 due soon too

“However, for setups where a monitor is maxed out at 240 Hz or 144 Hz, as many gaming panels are, using 6x MFG would be overkill. This is where Dynamic MFG comes into play. The technology determines which MFG multiplier is needed based on the display's refresh rate capability and the input framerate from the upscaler. NVIDIA calls this the "automatic transmission" for MFG, drawing a parallel to modern vehicle automatic transmission systems that switch gears based on demand. In graphically intensive scenarios, the multiplier can scale up to 4x, 5x, or 6x, while lighter scenes like settings menus or static sequences may only require a 2x multiplier to hit the target frame rate.

MFG 6x and Dynamic MFG remain exclusive to NVIDIA's GeForce RTX 50 Series "Blackwell" GPUs, which rely on hardware flip-metering unavailable on previous generations. On these GPUs, the jump from 4x to 6x Multi Frame Generation alone delivers up to a 35% increase in 4K frame rates in path-traced titles. The regular DLSS 4.5 Super Resolution technology, based on the second-generation Transformer model, is already available across over 400 games, though it is a visual improvement only with no direct performance boost. Below you can see dynamic frame generation parameters in the NVIDIA App with the DLSS Override settings.”

Source: https://www.techpowerup.com/347894/...rame-generation-and-6x-mode-officially-arrive

It may not make sense but I wish partial multipliers were a thing. If I can run a game at 100 FPS I'd love to use frame gen to hit the last 20 FPS to hit my cap of 120Hz, instead of dropping my real FPS to 60FPS to double it to 120 FPS and give me way more latency in the process. But yeah it doesn't work that way sadly

I'd need like a 360Hz monitor before I'd ever think of using 6x framegen. I get they say this feature would be nice even if you have 144Hz, but honestly I disagree. The difference in shifting from like a 2x multiplier to 3x is way bigger than shifting from 5x to 6x. This will only really look and feel smooth in the 4x-6x range

ivandagiant said:

It may not make sense but I wish partial multipliers were a thing. If I can run a game at 100 FPS I'd love to use frame gen to hit the last 20 FPS to hit my cap of 120Hz, instead of dropping my real FPS to 60FPS to double it to 120 FPS and give me way more latency in the process. But yeah it doesn't work that way sadly

I'd need like a 360Hz monitor before I'd ever think of using 6x framegen. I get they say this feature would be nice even if you have 144Hz, but honestly I disagree. The difference in shifting from like a 2x multiplier to 3x is way bigger than shifting from 5x to 6x. This will only really look and feel smooth in the 4x-6x range

Click to expand...

I like the fact it is dynamic though. I have a 144Hz panel, and when I played Doom TDA fully maxed with path tracing, 4X was smooth as butter, legit. Most games outside of that (if I even need it), I run 2x, maybe 3x. I kind of like the idea of set a frame cap and forget about it honestly.

View: https://youtu.be/iIvpt1CfKKI

erek said:

DLSS 5 due soon too

Click to expand...

Didn't they say they were shooting for fall? That doesn't seem that soon.

Looking forward to trying out dynamic frame gen.

wonder if this is it,


https://www.techpowerup.com/download/nvidia-dlss-3-frame-generation-dll/

ivandagiant said:

It may not make sense but I wish partial multipliers were a thing. If I can run a game at 100 FPS I'd love to use frame gen to hit the last 20 FPS to hit my cap of 120Hz, instead of dropping my real FPS to 60FPS to double it to 120 FPS and give me way more latency in the process. But yeah it doesn't work that way sadly

I'd need like a 360Hz monitor before I'd ever think of using 6x framegen. I get they say this feature would be nice even if you have 144Hz, but honestly I disagree. The difference in shifting from like a 2x multiplier to 3x is way bigger than shifting from 5x to 6x. This will only really look and feel smooth in the 4x-6x range

Click to expand...

I imagine it would be really hard to keep good frame pacing and do that. Not saying it would be impossible, but you'd have to figure out how and when to only generate more for certain frames rather than others.

as long as they use 2 frame to interpolate in between, non multiple with good frame pacing could be hard to do, at least that a feeling....

For the spatial reprojection frame (Reflex 2.0?) that seem more possible, those could make sure to always generate a frame for each of your monitor refresh, they would need to combine the 2, because right now this seem to need a good base frame rate/monitor ratio to work well (having a 600hz type maybe that easier)

ivandagiant said:

It may not make sense but I wish partial multipliers were a thing. If I can run a game at 100 FPS I'd love to use frame gen to hit the last 20 FPS to hit my cap of 120Hz, instead of dropping my real FPS to 60FPS to double it to 120 FPS and give me way more latency in the process. But yeah it doesn't work that way sadly

I'd need like a 360Hz monitor before I'd ever think of using 6x framegen. I get they say this feature would be nice even if you have 144Hz, but honestly I disagree. The difference in shifting from like a 2x multiplier to 3x is way bigger than shifting from 5x to 6x. This will only really look and feel smooth in the 4x-6x range

Click to expand...

Why? 100fps is more the enough for any game. I would rather run 60-100 fps of real fps then 120 with frame gen. If you are only getting 40-50 it probably makes sense to ues 2x mode but for some reason Nvidia is pushing this as we need 240+ fps of fakeness

tunatime said:

Why? 100fps is more the enough for any game. I would rather run 60-100 fps of real fps then 120 with frame gen. If you are only getting 40-50 it probably makes sense to ues 2x mode but for some reason Nvidia is pushing this as we need 240+ fps of fakeness

Click to expand...

Have you used it? The latency will basically match whatever your base frame rate latency is. So if you are already pushing 100 FPS, if you can use MFG and get more FPS, there is virtually no latency penalty, and it does feel smoother if you have a high refresh rate monitor.

If your base framerate is low, thus latency high, the latency penalty is still basically nothing, but it will still feel bad overall.

It is really a free boost if you have the ass to already run good frames and settings to allow more FPS basically "for free". It can really smooth things out, I use it all the time, even with a 5090.

III_Slyflyer_III said:

Have you used it? The latency will basically match whatever your base frame rate latency is. So if you are already pushing 100 FPS, if you can use MFG and get more FPS, there is virtually no latency penalty, and it does feel smoother if you have a high refresh rate monitor.

If your base framerate is low, thus latency high, the latency penalty is still basically nothing, but it will still feel bad overall.

It is really a free boost if you have the ass to already run good frames and settings to allow more FPS basically "for free". It can really smooth things out, I use it all the time, even with a 5090.

Click to expand...

Yep, I've described it often as a "win more" technology. It will not, as nVidia sometimes wants to imply, take a game running at 20fps and make it wonderful. It will take a game that is running at 60fps and make it smoother.

Everyone is going to have their own threshold but I personally find for most games so long as I can get in the 50s for FPS, it feels and looks good. Below that, problems can start happening and of course, higher base FPS is always better. With a decent base FPS I find the game can both look and feel better.

So for something like Cyberpunk 2077, which I'm replaying, 3x MFG works pretty good, though I'll be trying the dynamic mode now that it's out. I have it cranked and modded and with that, my base FPS varies from around 50 at the absolute lowest to 70ish at the absolute highest. So 3x MFG puts it in the 150-210 range which looks nice and smooth. For Hitman 3 with RT it generally is in the 90-110 base range, so for that 2x works good. Once a game gets much past 110, I'd rather disable it and just have native frame rate instead of capping it, though I'm sure it'll depend on the game. In neither case is it necessary, the game is perfectly playable without framegen, but it looks and feels better to me with framegen on.

This is on a 240Hz monitor. Now if I had a 500Hz monitor, well then I'd want higher multipliers. That is where it is getting mandatory if you want to use your monitor's refresh. Even for non-RT stuff that is not too graphically intense, I rarely see something cap out my 240Hz. So for those ultra high refresh ones, and the even ultra-higher ones that will no doubt come in the future, it becomes even more useful.

Sycraft said:

Yep, I've described it often as a "win more" technology. It will not, as nVidia sometimes wants to imply, take a game running at 20fps and make it wonderful. It will take a game that is running at 60fps and make it smoother.

Everyone is going to have their own threshold but I personally find for most games so long as I can get in the 50s for FPS, it feels and looks good. Below that, problems can start happening and of course, higher base FPS is always better. With a decent base FPS I find the game can both look and feel better.

So for something like Cyberpunk 2077, which I'm replaying, 3x MFG works pretty good, though I'll be trying the dynamic mode now that it's out. I have it cranked and modded and with that, my base FPS varies from around 50 at the absolute lowest to 70ish at the absolute highest. So 3x MFG puts it in the 150-210 range which looks nice and smooth. For Hitman 3 with RT it generally is in the 90-110 base range, so for that 2x works good. Once a game gets much past 110, I'd rather disable it and just have native frame rate instead of capping it, though I'm sure it'll depend on the game. In neither case is it necessary, the game is perfectly playable without framegen, but it looks and feels better to me with framegen on.

This is on a 240Hz monitor. Now if I had a 500Hz monitor, well then I'd want higher multipliers. That is where it is getting mandatory if you want to use your monitor's refresh. Even for non-RT stuff that is not too graphically intense, I rarely see something cap out my 240Hz. So for those ultra high refresh ones, and the even ultra-higher ones that will no doubt come in the future, it becomes even more useful.

Click to expand...

Totally agree. I would only play at a lower base FPS if it's something novel like Portal RTX where there was nothing like it at the time. Some sort of technical showcase you can't experience any other way.

But I'm typing this with a 5090 in my ivory tower.

There are a lot of people that game on budget hardware at sub 30 fps and just think that's normal, and it's not uncommon for console plebs to play with triple digit input lag. There are plenty of people that would use multiframe gen in a heartbeat.

sharknice said:

Totally agree. I would only play at a lower base FPS if it's something novel like Portal RTX where there was nothing like it at the time. Some sort of technical showcase you can't experience any other way.

But I'm typing this with a 5090 in my ivory tower.

There are a lot of people that game on budget hardware at sub 30 fps and just think that's normal, and it's not uncommon for console plebs to play with triple digit input lag. There are plenty of people that would use multiframe gen in a heartbeat.

Click to expand...

True, though I find it isn't just lag, at lower FPS you start to experience artifacting as well. The bigger the gap between frames, the more trouble framegen has making the generated frames look good. Also the more time your eyes are seeing them so the more issues will stand out. When I tried DLAA with Cyberpunk (just for fun), the base FPS was too low and though 4x MFG gets the FPS up to a fine amount, the experience wasn't good. Lag, but also I started noticing artifacts. So instead I run it on Balanced to have a higher base FPS.

Personally I'd turn down graphics until I was able to get something in the 50s (maybe 40s) base FPS. In fact I do just that on my laptop which does not have a 5090. It has a 40 series so only 2x frame gen but ya, I try to make sure I can get at least 50fps base/ 100fps doubled in a game and I'll tune my settings for that. It certainly lets me get them higher than if I wanted 100fps base though.

Either way, if it is something you have I encourage people to try it. I've heard the AMD version of framegen has some gremlins, particularly related to VRR, but the nVidia one works pretty well. If you have a game and it is running at half your monitor's max rate or less, flip it on and see. You might just like it. Won't work well for all games, I've seen some with issues. Hogwarts Legacy has extremely bad and noticeable UI issues when I tried (these are supposedly fixed now). It's worth trying though, maybe it gets you a smoother experience, or lets you turn up the graphics a bit more. If it doesn't do the trick for a given game, leave it off.

As an example of a game where I don't use it, Tempest Rising. Game has very simple graphics and if it isn't refresh limited is usually CPU limited. Even then it is running at like 180fps+. So no reason for it, even when it isn't maxing out the refresh rate I'd rather have the higher base FPS than limit it to 120 and use frame gen.

You can do whatever you like with it on a per-game basis.

I enabled it for Stalker 2, can't really tell any difference to be honest.

Garbage technology.

View: https://youtu.be/QBqF8FASl88?si=KQyUuXOrVuqL976D

DukenukemX said:

Garbage technology.

Click to expand...

Indeed, when you're intentionally breaking it like he does here. Esp. when he's clearly exceeding the VRAM (pegged at 8GB as shown on screen) on the CP2077 test with Path Tracing (RT Overdrive) enabled without FG, which defeats any point in testing FG or anything else for that matter.

An important note that he's ommiting here is that MFG performance and latency improves considerably on higher end GPUs with more tensor cores as well. So if you run these same tests on faster 50 series GPUs, there will be less latency added in those exact same tests.

Personally in the few games I've played that support MFG (CP2077, Portal RTX, RE9 recently) on my 5080, added latency was negligible at 2-4x with a base frame rate of 60+, which is what Nvidia recommends using FG with. I preferred having the added frames on my 240Hz display and I generally stay at 3x to stay near that cap. 4-6x MFG should be reserved for those with 300Hz+ displays, which are getting somewhat common now.

T4rd said:

Indeed, when you're intentionally breaking it like he does here. Esp. when he's clearly exceeding the VRAM (pegged at 8GB as shown on screen) on the CP2077 test with Path Tracing (RT Overdrive) enabled without FG, which defeats any point in testing FG or anything else for that matter.

An important note that he's ommiting here is that MFG performance and latency improves considerably on higher end GPUs with more tensor cores as well. So if you run these same tests on faster 50 series GPUs, there will be less latency added in those exact same tests.

Personally in the few games I've played that support MFG (CP2077, Portal RTX, RE9 recently) on my 5080, added latency was negligible at 2-4x with a base frame rate of 60+, which is what Nvidia recommends using FG with. I preferred having the added frames on my 240Hz display and I generally stay at 3x to stay near that cap. 4-6x MFG should be reserved for those with 300Hz+ displays, which are getting somewhat common now.

Click to expand...

One of the perceived benefits of frame generation is to help out those with weaker GPUs, which is obviously not working out that way. It's the reason why I still use a 1080P display, but I did upgrade to 100Hz. In order to drive something like a 4K 240Hz display would require a monster of a GPU, or a game that's well optimized. The problem is neither of these things easily exist and so we turn to technologies like upscaling and now frame generation, both of which cause their own fair share of problems. At 6X frame generation you will get ghosting and artifacting as well as increased latency. Couple this with upscaling which also ads ghosting, and these technologies will make playing games much worse. The way I look at it, a 4K 240Hz display is mostly meant for older games, which can easily achieve those frame rates at 4K.

T4rd said:

An important note that he's ommiting here is that MFG performance and latency improves considerably on higher end GPUs

Click to expand...

No, he explicitly said he chose the weakest GPU possible, and then he deliberately ran games with settings too high. That's kind of what he does. You can't draw any meaningful performance inferences from Dawid videos, except for when you're trying to use low-powered hardware. Duke knows this and ignores it. I've pointed it out multiple times.

1_rick said:

No, he explicitly said he chose the weakest GPU possible, and then he deliberately ran games with settings too high. That's kind of what he does. You can't draw any meaningful performance inferences from Dawid videos, except for when you're trying to use low-powered hardware. Duke knows this and ignores it. I've pointed it out multiple times.

Click to expand...

Oh yeah, I'm well aware of him and have fed him a few times with the bait he constantly throws out. He does the same thing in console threads, throwing out wildly skewed videos and comparisons to make them seem much weaker than they are or have zero advantages or use cases in the market. He's just a typical PCMR guy that burries his head in the sand when he's wrong or severely biased about anything.

T4rd said:

Personally in the few games I've played that support MFG (CP2077, Portal RTX, RE9 recently) on my 5080, added latency was negligible at 2-4x with a base frame rate of 60+, which is what Nvidia recommends using FG with. I preferred having the added frames on my 240Hz display and I generally stay at 3x to stay near that cap. 4-6x MFG should be reserved for those with 300Hz+ displays, which are getting somewhat common now.

Click to expand...

Ya this guy is just being a tool "I don't like this technology so let me intentionally break it and claim it sucks!" Personally, I think it is amazing. I can't believe how smooth it looks and how few problems I see. It just looks GOOD in games. Let's me really appreciate my 240Hz monitor. I've been replaying Cyberpunk, setting cranked up, graphics mods installed, and 3x MFG is letting me get framerates close to my monitor's max in most cases. Latency seems fine:



It's fuzzy on account of being a camera picture of an HDR screen (none of the screen caps I use capture the overlay) but it says 36.2ms latency. So, like, 1% of what he's showing. Net effect is the game feels good and looks extremely smooth. When the new update for dynamic framegen came out it actually broke framegen in Cyberpunk (just had to turn it off and on to make it work) and I was surprised at just how jerky it looked at 60-70 native FPS by comparison.

Now that said, nVidia does need to shut up about the "Ermehged this will take a 15-20FPS slideshow and make it smooth!" kind of marketing you see sometimes. It won't do a good job there. If you are getting 15 FPS native, and you run 6x FG that 90 FPS you get is going to feel crappy, and you'll see visual artifacts from the framegen to boot. But that doesn't mean it is a bad technology. Likewise it takes VRAM, so you need to have enough to run it well. That will depend on the game, your video card, and so on. However, if you are nearly maxing out VRAM with a game native, then no, framegen is not a good solution. You'd need to turn down graphics to free up memory first.

As I've said before, and I'll say a hundred times more: It is a win more technology. It isn't going to magically take 10FPS to 60FPS and thus unplayable to playable. What it will do is take 50FPS to 100, 150, 200, or more and take it from choppy to smooth. If you have a 60Hz monitor, ignore it, unlikely to do anything for you. I mean you can play with it for fun, but I don't imagine it'll be useful. If you have a 120-144Hz monitor, then 2x FG is something of interest. Any time your native rendering is falling below half of your max, maybe even a little more, see if it is worth using. Like on a 144Hz screen I might take 60 FPS doubled to 120 instead of 70 FPS native, and for sure I'd take it turned on at less than 60 FPS native. If you have a 240Hz screen then 3x becomes interesting, and maybe even 4x in some edge cases. If you have a 360Hz, 500Hz or more, well then that's where 5x and 6x become interesting. I mean shit, at 500Hz even if you have pretty high native FPS you still are going to want a lot of framegen. Like you could be playing Hitman 2 with RT off, which a powerful system can drive to 200fps+... and you still are less than half your max framerate, so 2x at least would be useful. Even well optimized games that are more intense could want for more. I mean say you have a game running in the 100-120 FPS range with graphics turned up. That's good performance, nothing to complain about, and on a 120Hz monitor you'd be real happy. However your 500Hz monitor is being wasted. Turn on FG though, and then 4x-5x is going to get you a smoother experience and use that fast refresh rate.

It's not magic, and it's not prefect, and it doesn't work well in every game, but damn if it isn't cool when it does work well. 100% worth trying if you have a card that supports it and a game that can use it. This includes 40 series owners. While you don't get the higher multipliers, 2x is still great.

1_rick said:

No, he explicitly said he chose the weakest GPU possible, and then he deliberately ran games with settings too high. That's kind of what he does. You can't draw any meaningful performance inferences from Dawid videos, except for when you're trying to use low-powered hardware. Duke knows this and ignores it. I've pointed it out multiple times.

Click to expand...

I did point out that he used a weak GPU and how this technology isn't useful for weak GPU's. Only person ignoring this is you.

DukenukemX said:

I did point out that he used a weak GPU and how this technology isn't useful for weak GPU's. Only person ignoring this is you.

Click to expand...

It's useful on weak gpus too, just with appropriate settings .

GoldenTiger said:

It's useful on weak gpus too, just with appropriate settings .

Click to expand...

Yep. While not a massively weak GPU, my laptop has a 4080 in it. That's 12GB VRAM and speeds a little slower than a desktop 4070. 2x framegen works well on it. I have to keep an eye on settings, of course, there are plenty of games that can overrun 12GB of VRAM but it works so long as you don't do that (same as with no framegen for that matter). It also has a 240Hz display, so framegen is a worthwhile thing to try for sure, as almost nothing is pushing it native with that GPU.

DukenukemX said:

I did point out that he used a weak GPU and how this technology isn't useful for weak GPU's. Only person ignoring this is you.

Click to expand...

Not really, you called it "garbage technology" because a dipstick Youtuber overloaded his GPU and then acted like it is framegen's fault. If you exceed a card's VRAM, framegen or no, everything goes to shit. System RAM is too slow, this is how it has always been.

That aside, let's say it only did work on higher end cards... still not garbage. There's nothing wrong with tech that makes the high end better.

Sycraft said:

Yep. While not a massively weak GPU, my laptop has a 4080 in it. That's 12GB VRAM and speeds a little slower than a desktop 4070. 2x framegen works well on it. I have to keep an eye on settings, of course, there are plenty of games that can overrun 12GB of VRAM but it works so long as you don't do that (same as with no framegen for that matter). It also has a 240Hz display, so framegen is a worthwhile thing to try for sure, as almost nothing is pushing it native with that GPU.

Click to expand...

What resolution are you running? I run a desktop RTX 4070S, and even using 2xFG I never get close to maxing out my vram at high/ultra settings @ 1200p.

Sycraft said:

Yep. While not a massively weak GPU, my laptop has a 4080 in it.

Click to expand...

That's not a weak GPU.

Not really, you called it "garbage technology" because a dipstick Youtuber overloaded his GPU and then acted like it is framegen's fault. If you exceed a card's VRAM, framegen or no, everything goes to shit. System RAM is too slow, this is how it has always been.

Click to expand...

You're paying attention to that part of the "dipstick Youtuber" when he overloads it. Even at 4x, he has higher frame latency than I have in network latency in World of Warcraft. He even says it feels like 30fps. At 6X he sees it jump to 50ms, which makes this garbage technology.

That aside, let's say it only did work on higher end cards... still not garbage. There's nothing wrong with tech that makes the high end better.

Click to expand...

This guy did a test with 6X frame generation and it doesn't get any better on a RTX 5080. If anything, Dawid Does Test makes DLSS 4.5 6X frame generation look less bad.

• Crimson Desert he get 61ms at 6X
• Cyberpunk 2077 gets 57ms at 6X
• Evil Requiem gets 55ms at 6X
• S.T.A.L.K.E.R. 2 gets 44ms at 6X
• Black Myth Wukong gets 57ms at 6X
• DOOM The Dark Ages gets 59ms at 6X
• Alan Wake 2 gets 56ms at 6X

Keep in mind that World of Warcraft for me has a network latency of 30ms. The input lag from DLSS4.5 6X is just horrible. Garbage technology.


View: https://youtu.be/2KohTjXImK0?si=4cO4loeW1SCaD41P

GoldenTiger said:

It's useful on weak gpus too, just with appropriate settings .

Click to expand...

DukenukemX said:

This guy did a test with 6X frame generation and it doesn't get any better on a RTX 5080. If anything, Dawid Does Test makes DLSS 4.5 6X frame generation look less bad.

• Crimson Desert he get 61ms at 6X
• Cyberpunk 2077 gets 57ms at 6X
• Evil Requiem gets 55ms at 6X
• S.T.A.L.K.E.R. 2 gets 44ms at 6X
• Black Myth Wukong gets 57ms at 6X
• DOOM The Dark Ages gets 59ms at 6X
• Alan Wake 2 gets 56ms at 6X

Click to expand...

Of course the reality is, such latency really doesn't matter in non competitive single player titles.

Mazzspeed said:

Of course the reality is, such latency really doesn't matter in non competitive single player titles.

Click to expand...

This is the same argument for cloud gaming. It's a bigger problem when you use a mouse because you can feel it lagging. People who use gamepads already have higher input lag, but a mouse with higher input lag feels horrible. A game like Expedition 33 won't be noticeable, until you get to the parts of the game where you have to perfect parry. It's a turn based combat game, but the parry mechanic is really tight timing. That 50ms is going to get you killed when you fight DLC bosses. Going back to shit youtuber, you can see the delay from his mouse movement and this is 4X. That's really bad.



View: https://youtu.be/YBMu9n49_S0?si=aLlSPb6JXiRjnnuc

Something is off with his wireless mouse, display, and/or settings (Reflex disbled somehow). I just fired up CP2077 to test again with 4x MFG and was around 35ms latency and my input was much more responsive than he's showing there on my wired G502 mouse and 5080. Absolutely no perceptable input delay for me at ~150 FPS/35ms latency showing in the Nvidia HUD while waving my mouse around like that in a vehicle.

Network latency isn't comparable to input latency either, so dunno why you're even bringing that up. Many games can have 50+ms input latency (RDR2 for example is pretty high) before any FG or other settings that would otherwise affect latency are applied. Also, many games can have the same or similar input latency with Reflex and FG enabled as non-Nvidia GPUs or consoles running the game without FG. Digital Foundry showed this in a couple of their videos testing it a while back.

I get being sceptical about it if you've never used it, which you haven't (at least on Nvidia). As I def was before I swapped my 3080 for a 5080. But after trying it, I generally prefer to play with it enabled as long as my base framerate is over 60 as precribed by Nvidia and AMD. It's not comparable to streaming games from the cloud or network latency at all and is virtually imperceptable from the native latency.

DukenukemX said:

This is the same argument for cloud gaming. It's a bigger problem when you use a mouse because you can feel it lagging.

Click to expand...

Cloud gaming is vastly different as ping becomes a problem when cloud gaming. I honestly believe those that claim to be sensitive to input lag on single player games are suffering from a form of Dunning-Kruger, furthermore, that video seems off.

/me periodic check in to DLSS thread.
/me sees same 4 people going back and forth.
/me goes back to gaming with DLSS.

4saken said:

/me periodic check in to DLSS thread.
/me sees same 4 people going back and forth.
/me goes back to gaming with DLSS.

Click to expand...

I love me some DLSS.

Mazzspeed said:

I love me some DLSS.

Click to expand...

“DLSS Enabler Can Now Simulate DLSS 4 Multi-Frame Gen X5 & X6 on Any DX12 GPU​

APRIL 12, 2026 JOHN PAPADOPOULOS 6 COMMENTS
A brand new version of DLSS Enabler has been released, allowing users to simulate DLSS 4 Multi-Frame Generation X5 and X6 on any DX12 GPU. This is something that may please a lot of users who wanted to try these MFG modes but were unable to as they are exclusive to the RTX 50 series GPUs.”

https://www.dsogaming.com/mods/dlss...dlss-4-multi-frame-gen-x5-x6-on-any-dx12-gpu/

erek said:

“DLSS Enabler Can Now Simulate DLSS 4 Multi-Frame Gen X5 & X6 on Any DX12 GPU​

APRIL 12, 2026 JOHN PAPADOPOULOS 6 COMMENTS
A brand new version of DLSS Enabler has been released, allowing users to simulate DLSS 4 Multi-Frame Generation X5 and X6 on any DX12 GPU. This is something that may please a lot of users who wanted to try these MFG modes but were unable to as they are exclusive to the RTX 50 series GPUs.”

https://www.dsogaming.com/mods/dlss...dlss-4-multi-frame-gen-x5-x6-on-any-dx12-gpu/

Click to expand...

Wouldn't this be somewhat similar to LSFG?

DukenukemX said:

That's not a weak GPU.

Click to expand...

It's pretty mid-range by desktop standards. Like I said, slightly weaker than a 4070. So not low end, but not a hitter. One generation old, 12GB RAM $600 launch MSRP. That's a mid-range unit. If your standard is "no that's STILL too powerful" well then I guess all I can say is "get better hardware." I have no issues with testing things on low end tech, but let's not pretend like nobody owns any mid-range or high end tech and that if it works well on those it isn't useful.

DukenukemX said:

You're paying attention to that part of the "dipstick Youtuber" when he overloads it. Even at 4x, he has higher frame latency than I have in network latency in World of Warcraft. He even says it feels like 30fps. At 6X he sees it jump to 50ms, which makes this garbage technology.

Click to expand...

What is it without framegen? You seem to mistake this for being a per-frame latency and think it should e super low. This is total engine latency, it is higher than the per-frame latency. You aren't going to see 4ms numbers, even if you have a 240Hz monitor, and they are aren't going to go up as much as you'd think with framegen.

Let's take Cyberpunk again. Here it is with no framegen:

84fps, 29.8ms total latency.

2x framegen:

147fps, 34.4ms total latency.

3x framegen (where I have been playing it):

213fps 34.4ms total latency.

4x framegen, which exceeds my monitor's refresh in this case:

267fps, 39.3ms total latency.

Also it should be noted because this is measuring engine latency, it bounces around a bit even when your FPS doesn't. I tried to get lowest latency in all pictures to keep it consistent. So even with no FG, with the settings I use, I'm looking at a minimum of about 30ms of engine latency. 2-3x FG add maybe 4-5ms more latency, a worthwhile tradeoff for the smoothness. For 4x not sure if that's inherent to the larger generation, or because I've gone beyond monitor refresh.

Seems pretty not-garbage and usable to me.


Ok but maybe that's just because Cyberpunk is too high latency a game, I'm playing with settings turned up too much, blah, blah, blah. Well, let's try a much simpler game, Tempest Rising. In fact, here I'll switch my monitor to 1080p480 mode, so there's higher refresh available. I'm running it with Epic quality settings, DLSS Quality mode (since DLSS is needed for framegen).

Here's no framegen:


210fps, 15.3ms. This is CPU limited at this point, something you often see with high framerates.

Here's 2x framegen:


394fps, 14.2ms latency. I imagine the latency drop isn't because of framegen, just the natural variation you see. In this case, most of the latency is on the CPU/engine side, not the GPU side.


So I dunno man, seem pretty useful to me. In a game like Cyberpunk, where I'm heavily GPU limited if I want all the shiny graphics and I can't push high FPS (it goes a lot lower in intense scenes) it lets me have much, MUCH more fluid graphics with only a maybe 5ms latency penalty. I'll take it, that is exceedingly hard to notice, the lower, choppier, framerate is much more noticeable. In Tempest Rising it wouldn't actually be useful on my system since I'd much rather play in 4k240 mode than 1080p480 mode but it shows where it would be useful on a high refresh rate monitor, even if you have a fast rendering time. My CPU is just pegged out, can't get anymore than 210fps doesn't matter how powerful the GPU is. So if I had a 500Hz monitor, well then that's where I'm stuck, going to a lower DLSS mode won't help... but frame gen will. Now I can get close to 400fps and in this case, being CPU limited, not even cost myself any extra latency.

As ever, not a magic technology, not for all games, not for all cases. But garbage? Nah, it works well and the latency penalty is often minimal.

T4rd said:

Something is off with his wireless mouse, display, and/or settings (Reflex disbled somehow). I just fired up CP2077 to test again with 4x MFG and was around 35ms latency and my input was much more responsive than he's showing there on my wired G502 mouse and 5080. Absolutely no perceptable input delay for me at ~150 FPS/35ms latency showing in the Nvidia HUD while waving my mouse around like that in a vehicle.

Click to expand...

This is entirely based on your opinion of course. The other games I linked had a base frame latency of 40ms on average, without the DLSS frame generation. A lot of modern games already have it bad to begin with so why make it worse?

I get being sceptical about it if you've never used it, which you haven't (at least on Nvidia). As I def was before I swapped my 3080 for a 5080. But after trying it, I generally prefer to play with it enabled as long as my base framerate is over 60 as precribed by Nvidia and AMD. It's not comparable to streaming games from the cloud or network latency at all and is virtually imperceptable from the native latency.

Click to expand...

It's not like AMD doesn't have a similar technology or you can just buy Lossless Scaling version. You don't need Nvidia to experience this. Nvidia's version is just cleaner, but still has the same problems like any other frame generator.

Sycraft said:

So I dunno man, seem pretty useful to me. In a game like Cyberpunk, where I'm heavily GPU limited if I want all the shiny graphics and I can't push high FPS (it goes a lot lower in intense scenes) it lets me have much, MUCH more fluid graphics with only a maybe 5ms latency penalty. I'll take it, that is exceedingly hard to notice, the lower, choppier, framerate is much more noticeable. In Tempest Rising it wouldn't actually be useful on my system since I'd much rather play in 4k240 mode than 1080p480 mode but it shows where it would be useful on a high refresh rate monitor, even if you have a fast rendering time. My CPU is just pegged out, can't get anymore than 210fps doesn't matter how powerful the GPU is. So if I had a 500Hz monitor, well then that's where I'm stuck, going to a lower DLSS mode won't help... but frame gen will. Now I can get close to 400fps and in this case, being CPU limited, not even cost myself any extra latency.

As ever, not a magic technology, not for all games, not for all cases. But garbage? Nah, it works well and the latency penalty is often minimal.

Click to expand...

2X isn't that bad but I still don't see why even enable it? Your frame rates are considered very acceptable to begin with. Especially Tempest Rising where going from 210 fps to 394 was probably not needed. Games that get good frame rates will do better with frame generation, but that kinda defeats it's purpose. Unless you really need that 240fps at 4K because monitor is being underutilized? Keep in mind the higher the frame generation, the worse the ghosting gets. There is a visual penalty for turning on 6X.

DukenukemX said:

It's not like AMD doesn't have a similar technology or you can just buy Lossless Scaling version. You don't need Nvidia to experience this. Nvidia's version is just cleaner, but still has the same problems like any other frame generator.

Click to expand...

Wrong. AMD's doesn't have a reflex like tech requirement and what they have is in a low single digit number of games. Nvidia has reflex as a requirement for framegen which significantly lowers the latency, and an ml model that has been trained better, lowering artifacts to virtually none. Lossless scaling framegen runs slower, has no latency mitigation, and runs on the gpu cores further adding latency. So yes, an Nvidia card is required to judge this.

DukenukemX said:

This is entirely based on your opinion of course. The other games I linked had a base frame latency of 40ms on average, without the DLSS frame generation. A lot of modern games already have it bad to begin with so why make it worse?

Click to expand...

Here's another opinion: Running FG with Nvidia Reflex here, any marginal reported change in latency is well and truely within the margin of statistical error - So essentially there's no change at all, certainly nothing that can be perceived and certainly nothing as exaggerated as in that linked video.

DukenukemX said:

This is entirely based on your opinion of course. The other games I linked had a base frame latency of 40ms on average, without the DLSS frame generation. A lot of modern games already have it bad to begin with so why make it worse?

Click to expand...

Because it makes the game look and feel better? It isn't like I can't, and don't try it both ways. Cyberpunk with details cranked up looks choppy to me without framegen. One of the downsides of OLED, I notice low FPS more than I did on my rather blurry MiniLED monitor. It looks MUCH smoother particularly at 3x and I find that I enjoy that more. So why NOT do it? It's not a competitive game, I'm not sweating over my ranking, I want to enjoy it and framegen makes it more enjoyable. Same reason I run Path Tracing or Ray Reconstruction. Game looks better with them on then off, but they hit FPS, and then frame gen brings back the smoothness.

If it looks better and plays better, why shouldn't I use it? Because some Youtuber has a fit about it? Because of some worshiping of absolute lowest latency for no reason? Because of hatred of new tech? If I feel the game is better with it on, then I'll turn it on.

DukenukemX said:

2X isn't that bad but I still don't see why even enable it? Your frame rates are considered very acceptable to begin with. Especially Tempest Rising where going from 210 fps to 394 was probably not needed.

Click to expand...

I don't turn it on in Tempest Rising, I used that just a test to show what happens in a game that already has a high base framerate. I'm not turning my monitor down to 1920x1080, which is what I have to do for 480Hz. While that does look smoother, I'd rather have the crisp 4k visuals. The point was to show it on a lower requirement, higher FPS, game. Cyberpunk represents the very heavy end of graphics, this represents the very light. Also to demonstrate that even then, you have engine latency and that framegen isn't going to cause massive problems with it. Also showing why it becomes a thing that is more interesting as monitors get faster. 500Hz 2.5k OLEDs are already a thing. I have no doubt with faster scaler chips, and DP 2.1, we'll see 4k monitors in the 480Hz+ range. This was to demonstrate that even if you are a "fuck nice graphics" kind of person you STILL run in to cases where you can't get ultra-high FPS. There are other limiting factors.


You seem to have a dogmatic opposition to the technology for some reason which is silly. It's an option for people to use if they want, and in a lot of cases, it is a good option.

GoldenTiger said:

Wrong. AMD's doesn't have a reflex like tech requirement and what they have is in a low single digit number of games. Nvidia has reflex as a requirement for framegen which significantly lowers the latency, and an ml model that has been trained better, lowering artifacts to virtually none. Lossless scaling framegen runs slower, has no latency mitigation, and runs on the gpu cores further adding latency. So yes, an Nvidia card is required to judge this.

Click to expand...

AMD's Anti-Lag isn't required because it only works on AMD hardware. AMD's FidelityFX Super Resolution can work on any GPU. If AMD made it a requirement then it would only work on AMD hardware, but you can run both AMD"s Super Resolution with AMD's Anti lag. Also yes, Lossless Scaling is worse but I did already say that Nvidia's implementation is cleaner.

Mazzspeed said:

Here's another opinion: Running FG with Nvidia Reflex here, any marginal reported change in latency is well and truely within the margin of statistical error - So essentially there's no change at all, certainly nothing that can be perceived and certainly nothing as exaggerated as in that linked video.

Click to expand...

Except we can actually measure it. Not only that but going from no DLSS 4.5 to DLSS 4.5 6X will increase your frame latency by 50% or even double it in some cases. So it isn't a margin of statistical error. Maybe you can't feel the difference, but again that depends on the person. I can see it in Dawid Does Tech video, but maybe my eyes are faster?

Sycraft said:

Because it makes the game look and feel better?

Click to expand...

But it actually doesn't. The image quality gets worse because there is artifacting and blurring, while also increasing input lag which means it should also feel worse. Going from 40ms to 60ms is going to feel like there's a rubber band in the controls.

If it looks better and plays better, why shouldn't I use it? Because some Youtuber has a fit about it? Because of some worshiping of absolute lowest latency for no reason? Because of hatred of new tech? If I feel the game is better with it on, then I'll turn it on.

Click to expand...

Making a lot of assumptions there. First, is it the shit Youtuber guy? Here's another. Satisfied yet? Probably not. Second, latency is king. The reason people go for higher frame rates is because of lower latency. So in my opinion, going for a higher frame rather that increases latency is counter productive. Third, the extra frames aren't artifact free. You are losing image quality in the process.

Also, it's not a hatred for new technology, it's a hatred for new tech that comes with down sides. If there were no downsides then I wouldn't have a problem. In some games like Borderlands 4, the downsides are minimal. On the other hand, Nvidia Reflex didn't effect it's already high latency of 50 ms, while turning on frame generation also didn't increase the latency much either, even at 6X. But this is more of a testament to how badly optimized Borderlands 4 is and not how little of an effect DLSS4.5 6X has on games. Same goes for Expedition 33 as it has a lot of latency before Reflex turned on.

View: https://youtu.be/r-DtgAdZWdc?si=zaiw5e2SU6MdanXh

You seem to have a dogmatic opposition to the technology for some reason which is silly. It's an option for people to use if they want, and in a lot of cases, it is a good option.

Click to expand...

Not all technology is created equal when it comes to what benefits consumers. Frame generation doesn't feel like a step forward. It's like taking one step forward while taking three steps back. To me, this is a telling sign that Nvidia doesn't want to make notable improvements to their graphic cards, and have instead turned to things like frame generation. This blame also falls on game developers and their lack of optimizing games. There should be no reason why modern games have such high frame latency to begin with.

If you haven't noticed, I said nothing bad about Nvidia's and AMD's Reflex and Anti Lag. All benefits and no disadvantages. I even know of LatecnyFlex which is an open source alternative. Latency is king, and anything that can reduce it is a big win for gamers.

DukenukemX said:

Except we can actually measure it. Not only that but going from no DLSS 4.5 to DLSS 4.5 6X will increase your frame latency by 50% or even double it in some cases. So it isn't a margin of statistical error. Maybe you can't feel the difference, but again that depends on the person. I can see it in Dawid Does Tech video, but maybe my eyes are faster?

Click to expand...

I can see it in that video, and there's something else going on there making latency look far more exaggerated than it really is - is there even a comparison with FG disabled in that video? I didn't see one. Furthermore, using Nvidia Reflex, latency isn't increased by 50%.