Framerate Isn't Good Enough: Latency Pipeline, Input Lag, Reflex, & Engineering Interview
video description
Date: 2024-02-10
Related videos
Comments and reviews: 19
Joppi1992
The level of gameplay where this has any real significance to a player, is really only at the very very top end. People playing as a hobby, even playing really well, will be limited a lot more by coordination issues within the brain such as reacting with a decided action and especially when needing to improvise that decision based on facing a situation you weren't prepared for.
Then there are manual compensation that people have done since the oldest versions of counter-strike, where you'd pull the trigger before the target is right within the scope, based on growing used to the latencies. Some even prepare mentally based on locating sounds before seeing the target(s).
There are also tactical decisions and improvised actions, like choosing what to throw when like for example flash bangs and so on, which isn't bound by thin margins of FPS and the small latency difference it can provide.
High FPS beyond a certain point is way overrated, where you're better off just enjoying better graphics, as a non-professional player.
reply
The level of gameplay where this has any real significance to a player, is really only at the very very top end. People playing as a hobby, even playing really well, will be limited a lot more by coordination issues within the brain such as reacting with a decided action and especially when needing to improvise that decision based on facing a situation you weren't prepared for.
Then there are manual compensation that people have done since the oldest versions of counter-strike, where you'd pull the trigger before the target is right within the scope, based on growing used to the latencies. Some even prepare mentally based on locating sounds before seeing the target(s).
There are also tactical decisions and improvised actions, like choosing what to throw when like for example flash bangs and so on, which isn't bound by thin margins of FPS and the small latency difference it can provide.
High FPS beyond a certain point is way overrated, where you're better off just enjoying better graphics, as a non-professional player.
reply
explodatedfaces
I think the main problem that latency introduces is that the numbers may be smaller when you measure them than FPS will be, but they are even more meaningless. How many people can tell the difference between 27 milliseconds and 33 milliseconds How many people can feel the difference between when they click their mouse and when something happens and actually be right that it changed from last time As a general rule when you start to count fractions of a second less than 1/5 you start to show the biggest problem that the industry has; they are making products better over time, but failing to provide any need for those better products. I'll freely admit I cannot tell the difference in latency between a 3060 laptop and a 4090 on desktop. I can't identify the difference in latency between any two modern computers at all because the actual latency end to end is already so low that it takes my brain orders of magnitude longer to register any information.
reply
I think the main problem that latency introduces is that the numbers may be smaller when you measure them than FPS will be, but they are even more meaningless. How many people can tell the difference between 27 milliseconds and 33 milliseconds How many people can feel the difference between when they click their mouse and when something happens and actually be right that it changed from last time As a general rule when you start to count fractions of a second less than 1/5 you start to show the biggest problem that the industry has; they are making products better over time, but failing to provide any need for those better products. I'll freely admit I cannot tell the difference in latency between a 3060 laptop and a 4090 on desktop. I can't identify the difference in latency between any two modern computers at all because the actual latency end to end is already so low that it takes my brain orders of magnitude longer to register any information.
reply
stevekristoff4365
Glad that you're starting to look into this. Though this is MUCH larger than just for graphics. I.e. a very good example is storage (nvme/ssd/hdd); memory access; as well as general interrupt handling delays. The biggest item I run into pretty much daily is data subsystem performance (drive I/O basically). This is made worse as basically OEM's and all reviewers seem to be fascinated by 'streaming throughput' to a subsystem when this workload basically is completely non-existent for any real world workload. (anyone who doubts this, load up diskmon and check for yourself, it doesn't exist). Wish you can add similar level of detail to any reviews you do with these other subsystems.
reply
Glad that you're starting to look into this. Though this is MUCH larger than just for graphics. I.e. a very good example is storage (nvme/ssd/hdd); memory access; as well as general interrupt handling delays. The biggest item I run into pretty much daily is data subsystem performance (drive I/O basically). This is made worse as basically OEM's and all reviewers seem to be fascinated by 'streaming throughput' to a subsystem when this workload basically is completely non-existent for any real world workload. (anyone who doubts this, load up diskmon and check for yourself, it doesn't exist). Wish you can add similar level of detail to any reviews you do with these other subsystems.
reply
facksmasheen
I have to once again humbly beg for you to take a look at Tekken 8. Specifically with regard to latency and frametime consistency. It's a unique situation with the game being hard locked to 60fps and more graphics settings than ever before, it's hard to know what settings have a positive effect on those metrics. Sometimes I will change a setting and I'm not sure if it's better or worse- is it a placebo effect when it feels better Is it actually better, but I'm used to the old laggier settings I suppose I could use RivaTuner and experiment but I don't trust that I have the knowhow to not just waste hours of my time. Thanks Steve :)
reply
I have to once again humbly beg for you to take a look at Tekken 8. Specifically with regard to latency and frametime consistency. It's a unique situation with the game being hard locked to 60fps and more graphics settings than ever before, it's hard to know what settings have a positive effect on those metrics. Sometimes I will change a setting and I'm not sure if it's better or worse- is it a placebo effect when it feels better Is it actually better, but I'm used to the old laggier settings I suppose I could use RivaTuner and experiment but I don't trust that I have the knowhow to not just waste hours of my time. Thanks Steve :)
reply
Cylonknight
All of this in this video is why in any game I turn off all DLSS, all ambient occlusion, make it full screen when applicable, chromatic aberration, a few other settings but uh, I'm at work right now and can't remember off the top of my head . There is a distinctive difference between feeling the movement of your character in a game engine with those settings changed. I equate it to feeling the road through a car, some people do and some just don't understand it. Specially in FPS games, like siege, you'll always want ambient occlusion off, it's a game where it's really easy to tell the difference between input latency.
reply
All of this in this video is why in any game I turn off all DLSS, all ambient occlusion, make it full screen when applicable, chromatic aberration, a few other settings but uh, I'm at work right now and can't remember off the top of my head . There is a distinctive difference between feeling the movement of your character in a game engine with those settings changed. I equate it to feeling the road through a car, some people do and some just don't understand it. Specially in FPS games, like siege, you'll always want ambient occlusion off, it's a game where it's really easy to tell the difference between input latency.
reply
cyberwolf575
I know this is a game you probably won't include because it's so hard to reproduce the same results in it, but Escape From Tarkov is a perfect example of input latency vs frame rate. The game is so weirdly optimized that sometimes getting 140fps you feel like you are getting 30FPS as far as input latency goes, and at other times getting 60fps feels smooth as hell, and legit more smooth than some other games do at 200 fps. That game legit makes no sense. But from a purely scientific perspective, I always wondered if someone could look into it and figure it out.
reply
I know this is a game you probably won't include because it's so hard to reproduce the same results in it, but Escape From Tarkov is a perfect example of input latency vs frame rate. The game is so weirdly optimized that sometimes getting 140fps you feel like you are getting 30FPS as far as input latency goes, and at other times getting 60fps feels smooth as hell, and legit more smooth than some other games do at 200 fps. That game legit makes no sense. But from a purely scientific perspective, I always wondered if someone could look into it and figure it out.
reply
MultiZymethSK
This is the first time I understood what Reflex is actually doing, I think. If I understand it's trying to delay the CPU processing by the GPU render time minus CPU processing time, so that the CPU finishes processing just as the GPU gets freed to render another frame. I don't really know where the slightly lower GPU utilization comes from, though. I guess if you delay the CPU too much the GPU will not be busy 100% of the time, but then aren't you getting worse latency So shouldn't 100% GPU usage be the goal in this case
reply
This is the first time I understood what Reflex is actually doing, I think. If I understand it's trying to delay the CPU processing by the GPU render time minus CPU processing time, so that the CPU finishes processing just as the GPU gets freed to render another frame. I don't really know where the slightly lower GPU utilization comes from, though. I guess if you delay the CPU too much the GPU will not be busy 100% of the time, but then aren't you getting worse latency So shouldn't 100% GPU usage be the goal in this case
reply
MixesPWLSAW
Playing on Series S on Warzone last year, it was really bad input lag wise at 120hz refresh rate with a 240hz freesync monitor using HDMI 2.0 with a 10ms ping to the server.
In August using the same Xbox controller with it's 250mhz polling rate on a Nvidia 3060 12GB, with Nvidia Reflex on, it's like night and day with the same monitor and internet connection at 10ms to the server.
Also now seeing this with FSR 3 with Starfield, adding a lot of input lag into it, with the most recent patch compared to DLSS.
reply
Playing on Series S on Warzone last year, it was really bad input lag wise at 120hz refresh rate with a 240hz freesync monitor using HDMI 2.0 with a 10ms ping to the server.
In August using the same Xbox controller with it's 250mhz polling rate on a Nvidia 3060 12GB, with Nvidia Reflex on, it's like night and day with the same monitor and internet connection at 10ms to the server.
Also now seeing this with FSR 3 with Starfield, adding a lot of input lag into it, with the most recent patch compared to DLSS.
reply
De_kaid
That final bit about Reflex doesn't really make sense to me, if the system is gpu bound at 100 fps both times and you want to push 100 fps at all times then surely you'd always have to queue some frames as the gpu isn't gonna keep up its maximum all the time, Reflex isn't gonna magically up your GPU performance.
In the same way, wouldn't simply limiting the framerate to something slightly below your lowest gpu bound frametime be an even smoother experience as you'd have even less frame time fluctuation
reply
That final bit about Reflex doesn't really make sense to me, if the system is gpu bound at 100 fps both times and you want to push 100 fps at all times then surely you'd always have to queue some frames as the gpu isn't gonna keep up its maximum all the time, Reflex isn't gonna magically up your GPU performance.
In the same way, wouldn't simply limiting the framerate to something slightly below your lowest gpu bound frametime be an even smoother experience as you'd have even less frame time fluctuation
reply
im18yearsolder
I love this kind of content because it highlights stuff that is commonly overlooked by the majority of people. I noticed after getting my 4090 that the latency in gaming has been greatly reduced and whilst I didn't think much of it at first, it has been a noticeable improvement for me in games like Overwatch. Thanks Steve and team for this awesome educational content, it's always amazing to hear the experts from Nvidia talk because they know their shit.
reply
I love this kind of content because it highlights stuff that is commonly overlooked by the majority of people. I noticed after getting my 4090 that the latency in gaming has been greatly reduced and whilst I didn't think much of it at first, it has been a noticeable improvement for me in games like Overwatch. Thanks Steve and team for this awesome educational content, it's always amazing to hear the experts from Nvidia talk because they know their shit.
reply
JosefSpjut
One clarification to the answer to Steve's question at around 22:30. The input sampling is a uniform random variable with an average time of 5 ms in that example. This is because as Guillermo explained earlier, the input could have arrived anywhere from 0 to 10 ms before the frame, so the sampling causes an average of 5 ms of latency, though any particular frame has a roughly equal probability of being any value from 0 to 10 ms.
reply
One clarification to the answer to Steve's question at around 22:30. The input sampling is a uniform random variable with an average time of 5 ms in that example. This is because as Guillermo explained earlier, the input could have arrived anywhere from 0 to 10 ms before the frame, so the sampling causes an average of 5 ms of latency, though any particular frame has a roughly equal probability of being any value from 0 to 10 ms.
reply
mikeramos91
I know that more fps means lower frame time/latency but what I never understood was that if your gpu can do like 150fps but you decide to lock it at 60 you’re frame time/latency goes up. Normally the frame time/latency would be high due to the gpu struggling to produce frames, but if the gpu doesn’t struggle & it clearly has enough power for way more fps the frame time shouldn’t change if you decide to lock the fps
reply
I know that more fps means lower frame time/latency but what I never understood was that if your gpu can do like 150fps but you decide to lock it at 60 you’re frame time/latency goes up. Normally the frame time/latency would be high due to the gpu struggling to produce frames, but if the gpu doesn’t struggle & it clearly has enough power for way more fps the frame time shouldn’t change if you decide to lock the fps
reply
astreakaito5625
It's also not good enough for motion clarity, Gamers Nexus need to make a video or two about strobing or Black Frame Insertion features in certain monitors and TVs.
Because why bother with frame generation when BFI is literally better and boost motion clarity by more than x2.5 with practically no downsides I can't play videogames without BFI now, but most people don't even know about it.
reply
It's also not good enough for motion clarity, Gamers Nexus need to make a video or two about strobing or Black Frame Insertion features in certain monitors and TVs.
Because why bother with frame generation when BFI is literally better and boost motion clarity by more than x2.5 with practically no downsides I can't play videogames without BFI now, but most people don't even know about it.
reply
traitoR142
People are getting so anal it's honestly gotten ridiculous, getting hundreds and hundreds of frames per second now with ridiculous levels of graphic fidelity and still whining about the tiniest things, once upon a time people were happy if they could get 480P 30 fps and complained less. It's like people have no idea how lucky they are to have what they do now, just complain about everything.
reply
People are getting so anal it's honestly gotten ridiculous, getting hundreds and hundreds of frames per second now with ridiculous levels of graphic fidelity and still whining about the tiniest things, once upon a time people were happy if they could get 480P 30 fps and complained less. It's like people have no idea how lucky they are to have what they do now, just complain about everything.
reply
astrojunkey
Been waiting on a video like this! If anyone is interested in knowing more about polling (specifically on old consoles), there is a video called You will never see a Bomb Torizo skip in a World Record Super Metroid speedrun (NTSC) which talks about issues emerging at polling rates around 60Hz. It's very focused on Super Metroid, but I found it very interesting.
reply
Been waiting on a video like this! If anyone is interested in knowing more about polling (specifically on old consoles), there is a video called You will never see a Bomb Torizo skip in a World Record Super Metroid speedrun (NTSC) which talks about issues emerging at polling rates around 60Hz. It's very focused on Super Metroid, but I found it very interesting.
reply
thatguyat
this kind of testing is really important, recently i have been trying out frame generation in a bunch of games thanks to Lossless Scaling's implementation, and every single time i have decided against frame gen due to input latency. Frame gen from 85 FPS to 170 FPS feels so much worse than 90-130 real FPS with equivalent settings.
reply
this kind of testing is really important, recently i have been trying out frame generation in a bunch of games thanks to Lossless Scaling's implementation, and every single time i have decided against frame gen due to input latency. Frame gen from 85 FPS to 170 FPS feels so much worse than 90-130 real FPS with equivalent settings.
reply
ManuFortis
This video is straight up homework for basically every single online game player out there that complains about their setup, the network, other players, etc. Sorry if it causes ya any problems Steve and crew, but i'm gonna be copy and pasting the link to this video probably any time some uppity snot opens their mouth.
reply
This video is straight up homework for basically every single online game player out there that complains about their setup, the network, other players, etc. Sorry if it causes ya any problems Steve and crew, but i'm gonna be copy and pasting the link to this video probably any time some uppity snot opens their mouth.
reply
saricubra2867
99% of latency for a competitive shooter is your internet connection and monitor.
I play on a old 19 inch CRT at 1600 x 1200 67Hz or 1600x900 at 85Hz.
Counter Strike 2 is all about peaking (combined with proper grenade throws) CRTs have perfect motion clarity and they are analog, they don't add latency.
reply
99% of latency for a competitive shooter is your internet connection and monitor.
I play on a old 19 inch CRT at 1600 x 1200 67Hz or 1600x900 at 85Hz.
Counter Strike 2 is all about peaking (combined with proper grenade throws) CRTs have perfect motion clarity and they are analog, they don't add latency.
reply
AndrewFleites
Awesome! I've been wanting a deep dive into the importance of latency. It seems that 4k gamers are missing out on a lot of responsiveness, so the new oled monitors that can switch between 240hz 4k and 480hz 1080p should be a godsend. Now I'm just waiting for a reflex oled monitor at 1080p . . . one can dream.
reply
Awesome! I've been wanting a deep dive into the importance of latency. It seems that 4k gamers are missing out on a lot of responsiveness, so the new oled monitors that can switch between 240hz 4k and 480hz 1080p should be a godsend. Now I'm just waiting for a reflex oled monitor at 1080p . . . one can dream.
reply
Add a review, comment
Other channel videos
