Support NeoGAF

[Durante]

Here is the input lag study. It's part of a larger "G-sync 101" series of articles.

Originally posted by Paragon in the G-sync thread, I believe this deserves more recognition.

It's one of the most comprehensive studies of true in-game end-to-end (input-to-monitor) studies I've ever seen. The methodology is described in detail in the article, but basically it involves recording 1000 FPS video, having a specifically modified input device which lights a LED on input, and then counting the frames (which are ms) until a reaction occurs on screen.

There is a lot of great data in the article, but what I found particularly valuable is this:

What it shows are two important facts:

1. that G-sync allows you to set a frame limit slightly below your refresh rate while maintaining a smooth picture, which can massively reduce your input lag
2. that higher refresh rates improve your input behaviour considerably even if you don't match their refresh rate in terms of FPS

These are points I've argued for before, but it's good to have solid data backing up the theory. (Honestly, the difference in this particular game is in fact more pronounced than I would have expected)

 

[Son of a Witch]

Amazing article, now we just need a FreeSync comparison.

 

[mullet2000]

Interesting. I didn't realize having the cap way higher than your refresh rate was so bad.

I usually cap things at 140 on my 144hz/Gsync monitor, because I've heard before that capping a few frames below max is the best thing to do. Seems like this is good evidence that that's the way to go?

 

[Durante]

Interesting. I didn't realize having the cap way higher than your refresh rate was so bad.

I usually cap things at 140 on my 144hz/Gsync monitor, because I've heard before that capping a few frames below max is the best thing to do. Seems like this is good evidence that that's the way to go?

Yes, absolutely. That seems to be a constant throughout their testing.

 

[LelouchZero]

Wow this is fantastic, there's really good information here and the testing methodology is excellent!

It's very interesting to see that there's lower input latency at higher refresh rates, even at frame rates which don't match the refresh rate with G-sync.

Paragon and Durante, thanks for posting this!

 

[Cannon Goose]

Great article, I'd been looking forward to it for a while after reading their forums.

I found it really interesting that they got no input lag differences between full screen and windowed mode. Other tests I've seen have shown that windowed was always laggier.

 

[spootime]

Yes, absolutely. That seems to be a constant throughout their testing.

Does this apply to non-gsync monitors as well? People are often running CSGO uncapped (300+ frames) on 144hz monitors.

 

[Durante]

Does this apply to non-gsync monitors as well? People are often running CSGO uncapped (300+ frames) on 144hz monitors.

If you don't have VRR, then capping slightly below refresh rate will give you really bad consistent microstutter, so it's not really an option.

 

[Nokterian]

Very cool article indeed, went through my settings did the optimal ones and i have to test it yet but so far this one comprehensive test. I knew before that it should be below my refresh rate to keep it stable now i know 141fps is the limit to run it butter smooth.

 

[grendelrt]

Awesome article, does he say how he is limiting FPS? Overwatch has it built in, but outside of that is Rivatuner still the best way?

 

[mullet2000]

Awesome article, does he say how he is limiting FPS? Overwatch has it built in, but outside of that is Rivatuner still the best way?

Riva tuner still seems to be the best option, yeah.

 

[Duebrithil]

Quite meaty. Thanks for posting this guys.

 

[LumpOfCole]

This is why I love it when games let me set a specified framerate limit. I have a Predator X34 that has 100hz Gsync, and it's best for me to set games like Overwatch to a framerate cap of 96fps so that there's absolutely no tearing at all. I can't imagine even going back to 60hz permanently - a power outage threw off my Gsync to get stuck at 60hz until a hard reset and it was terrible.

 

[Arkanius]

TL;DR people:

Gsync On + Vsync On + Framelimit -3fps below your max

Ex: 144hz monitor, you should cap your framerate at 141

This gives you the stutter free, no tearing and no input lag experience.

 

[jim2point0]

This makes me sad that I don't have a gsync monitor yet.

 

[Mohasus]

Awesome article, does he say how he is limiting FPS? Overwatch has it built in, but outside of that is Rivatuner still the best way?

The image says in-game Limiter.

I remember another test that showed RTSS adding some lag, but I can't find it now.

 

[Darktalon]

What is better in overwatch , playing 141hz cap and gsync, or 120hz ULMB. And if it's ULMB, what should the cap be?

 

[creatchee]

Studies like this are going to inadvertently sell a lot of HDMI 2.1 TV's to gamers.

 

[grendelrt]

The image says in-game Limiter.

I remember another test that showed RTSS adding some lag, but I can't find it now.

I think Overwatch is the only game I currently play with it built in though, so have to do it somehow outside of that.

 

[Durante]

What is better in overwatch , playing 141hz cap and gsync, or 120hz ULMB. And if it's ULMB, what should the cap be?

Which one is better would seem to depend on whether you have a preference for slightly lower input lag (141+G-sync) or slightly less blur (120+ULMB).

Of course, with either choice, blur and input lag are already greatly reduced compared to a standard display.

 

[DouglasteR]

But Gsync + Vsync wouldnt annihilated themselves (effect wise) ?

 

[TylerDurden4321]

Wait, am I reading this right??

Even if the FPS are below the refresh rate, Fast Sync(right bottom) is still very superior to normal V-Sync(left top)?
Wasn't it advertised as "use this, if you have ~3x FPS of your display refresh rate" ??

 

[Dave_6]

TL;DR people:

Gsync On + Vsync On + Framelimit -3fps below your max

Ex: 144hz monitor, you should cap your framerate at 141

This gives you the stutter free, no tearing and no input lag experience.

Thank you! Though my rig won't hit my current cap of 144hz, I'm still going to change the cap in Rivatuner to 141.

 

[nextgeneration]

For the people who have 144 hz montiors, do you notice a big difference between 144 fps and 141 fps? Is there really that big of a difference?

 

[mullet2000]

For the people who have 144 hz montiors, do you notice a big difference between 144 fps and 141 fps? Is there really that big of a difference?

You will absolutely not be able to perceive the difference.

 

[LumpOfCole]

For the people who have 144 hz montiors, do you notice a big difference between 144 fps and 141 fps? Is there really that big of a difference?

No. I do however notice the occasional tearing that happens when V-Sync is off and the refresh rate / framerate is absolutely maxed. It's objectively better to bring that max framerate down by 3.

 

[Kolle]

that higher refresh rates improve your input behaviour considerably even if you don't match their refresh rate in terms of FPS

I have a g-sync monitor but I thought that the display matches the frame rate you get, so if you get for example 72 fps the refresh rate would be 72hz. How can the input lag then still be lower?

 

[Darktalon]

I have a g-sync monitor but I thought that the display matches the frame rate you get, so if you get for example 72 fps the refresh rate would be 72hz. How can the input lag then still be lower?

I believe what actually happens is that the monitor still prepares them just as fast as if it was 144hz, but delivers them evenly spaced to the frame rate its at. So you have 72 frames displayed but can read them out faster so it's not one of the bottlenecks.

 

[TheExodu5]

If you don't have VRR, then capping slightly below refresh rate will give you really bad consistent microstutter, so it's not really an option.

Without any vsync whatsoever, uncapped still offers the lowest input lag (even moreso than vsync) since the latest frame or partial frame will always be drawn at every polling interval, rather than waiting in a buffer for any additional period of time.

Non-vsync input lag should fluctuate between 1/framerate to 1/refreshrate (the input lag will vary across the screen as partial frames are being displayed).

Gets a little harder when we look at comparing 1/144Hz gsync to 1/120 ULMB (vsync off). Assuming you cap your framerate just below 144fps, the frame will stay in the buffer for a very short period of time, so you'd be looking at nearly 1/144Hz input lag. With 120Hz non-vsync, the input lag would again vary across the screen, ranging from 1/framerate to 1/120Hz at any given time. If we average it out, I would guess that you need to maintain higher than 168fps to have a lower effective input lag compared to 144Hz gsync.

 

[nextgeneration]

No. I do however notice the occasional tearing that happens when V-Sync is off and the refresh rate / framerate is absolutely maxed. It's objectively better to bring that max framerate down by 3.

You will absolutely not be able to perceive the difference.

Hmm, ok, well, I went ahead and limited my frame rate to 141 in rivatuner just for shits and giggles, since this is what is recommended - lol.

 

[Cels]

why does this happen? what is the gsync ceiling they're talking about here?




this one is also interesting, vsync off has a very tiny edge over vsync on..i thought it would be more of a difference

explained by this

the higher the refresh rate of the display, the faster the scanout speed becomes. This also explains why V-SYNC OFF's input lag advantage, especially at the same framerate as G-SYNC, is reduced as the refresh rate increases; single frame delivery becomes faster, and V-SYNC OFF has less of an opportunity to defeat the scanout.

 

[Arkanius]

why does this happen? what is the gsync ceiling they're talking about here?

The Gsync ceiling is at 144fps (to match the 144hz)
Notice how the input lag is great if the FPS limit is at 300 FPS, which means you are hitting the Gsync Limit at 144FPS, thus adding input lag.

When they FPS limit to below the ceiling (143, 142, 134) the Input lag is GREATLY reduced.

 

[Kolle]

I believe what actually happens is that the monitor still prepares them just as fast as if it was 144hz, but delivers them evenly spaced to the frame rate its at. So you have 72 frames displayed but can read them out faster so it's not one of the bottlenecks.

Yeah I think you're right, just got to this part in the article regarding 60 fps at 144hz where they state: "And while each frame is still rendered in 16.6ms, and delivered in intervals of 60 per second on the higher refresh rate display, they are scanned in at a much faster 6.9ms per."

 

[Knurek]

How does limiting the framerate to 120 fps on a 144 Hz G-Sync screen work? The article states setting the limit too low is a bad idea, but what exactly is 'too low'?
(My monitor has a firmware bug with some pixel transitions at anything above 120 fps)

 

[bee]

great read!

i got that gsync 2-3fps lower trick from nvidia inspector a while back when they labelled their frame rate limiter accordingly for 58 and 118fps, on 4k 60hz gsync screens the reduction in input lag is very noticeable indeed and actually makes it close to no gsync/vsync

 

[TheExodu5]

Okay, they also confirm that triple buffer vsync adds an extra frame of lag compared to double buffer.

My old tag used to be: will use d3doverrider to enable triple buffering instead of complaining about mouse lag in every PC game thread ever.

Finally...vindicated.

 

[Izuna]

This article saved my life.

 

[Lostconfused]

Okay, they also confirm that triple buffer vsync adds an extra frame of lag compared to double buffer.

My old tag used to be: will use d3doverrider to enable triple buffering instead of complaining about mouse lag in every PC game thread ever.

Finally...vindicated.

Doesn't actually say anything about triple buffer for applications not involving gsync.

Also hilarious that you still feel ass blasted about that tag.

 

[TheExodu5]

Doesn't actually say anything about triple buffer for applications not involving gsync.

Also hilarious that you still feel ass blasted about that tag.

Actually, they do mention it in passing.

As for triple buffer V-SYNC, it typically introduces an additional frame of delay over double buffer V-SYNC when the framerate exceeds the maximum refresh rate, and won’t be included in results due to the fact that G-SYNC is based on a double buffer, and when the rarely supported triple buffer option is available for use, it actually has two entirely separate methods, dependent in part on whether the game engine is based on OpenGL or DirectX.

I kept saying this for the longest time here but it was commonly accepted at GAF that Tripple Buffering reduced input lag over Double Buffering...which doesn't make sense. Triple buffering would only reduce input lag if the framerate were to fall below the monitor's refresh rate.

 

[Optimus Prime]

I see a lot of g-sync plus v-sync testing here, though the input lag is the same or better with v-sync off. Is that because v-sync (if enabled) kicks in outside the range of g-sync?

If my fps stay within the limit of my adaptive sync range (I'm using freesync on a 48-144hz monitor) by using a high enough in-game fps cap (142 in Rocket League), do I need to bother with v-sync?

Would there likely be any substantial difference if freesync was substituted here for g-sync in the testing?

I use freesync with in-game fps cap of 142 in Rocket League. No v-sync. Can get 170-210 fps if I set the cap to 250, though I'm sure I could get higher if I turned off some settings. So I'm wondering if I should keep my settings or if the higher framerate would mitigate the input lag but still give me the better responsiveness of higher framerates. And if I did this, should I turn off freesync or does it matter?

 

[Darktalon]

I see a lot of g-sync plus v-sync testing here, though the input lag is the same or better with v-sync off. Is that because v-sync (if enabled) kicks in outside the range of g-sync?

If my fps stay within the limit of my adaptive sync range (I'm using freesync on a 48-144hz monitor) by using a high enough in-game fps cap (142 in Rocket League), do I need to bother with v-sync?

Would there likely be any substantial difference if freesync was substituted here for g-sync in the testing?

I use freesync with in-game fps cap of 142 in Rocket League. No v-sync. Can get 170-210 fps if I set the cap to 250, though I'm sure I could get higher if I turned off some settings. So I'm wondering if I should keep my settings or if the higher framerate would mitigate the input lag but still give me the better responsiveness of higher framerates. And if I did this, should I turn off freesync or does it matter?

Without vsync you get torn frames. Do you want less input lag with torn frames, or minimal input lag with perfect displayed frames?
If freesync works like gsync, turn vsync on in control panel, cap framerate to 3 below monitors max refresh rate, and turn off vsync in all your games in-game settings.

 

[riflen]

I see a lot of g-sync plus v-sync testing here, though the input lag is the same or better with v-sync off. Is that because v-sync (if enabled) kicks in outside the range of g-sync?

Yes.

If my fps stay within the limit of my adaptive sync range (I'm using freesync on a 48-144hz monitor) by using a high enough in-game fps cap (142 in Rocket League), do I need to bother with v-sync?

No.

Would there likely be any substantial difference if freesync was substituted here for g-sync in the testing?

Don't know. Principles are the same but obviously these are tests on specific equipment.

I use freesync with in-game fps cap of 142 in Rocket League. No v-sync. Can get 170-210 fps if I set the cap to 250, though I'm sure I could get higher if I turned off some settings. So I'm wondering if I should keep my settings or if the higher framerate would mitigate the input lag but still give me the better responsiveness of higher framerates. And if I did this, should I turn off freesync or does it matter?

You will get tearing if the rate exceeds 144 fps. So it's a personal choice. AMD added the option to let the frame rate exceed refresh rate in FreeSync mode to give players the choice to get the lowest input latency at the expense of image stability.
The question is, is the input latency at 142 fps and 144 hz causing you problems? In this article on this equipment, that input latency is ~20 ms.

 

[Durante]

I kept saying this for the longest time here but it was commonly accepted at GAF that Tripple Buffering reduced input lag over Double Buffering...which doesn't make sense. Triple buffering would only reduce input lag if the framerate were to fall below the monitor's refresh rate.

This is not the case, and that part of the article isn't very clear.
It's right in that there are two entirely different things commonly called "triple buffering": one of them adds lag, one of them reduces it, and only one of them is what I would call triple buffering.

And in fact, that -- as in, real triple buffering -- is what NV has now branded FastSync! So with that you can actually use it in most games, and they do in fact test it in the article

 

[Felix Lighter]

Is RTSS the preferred method to cap framerate universally?

 

[Durante]

Is RTSS the preferred method to cap framerate universally?

The preferred method is to use an in-game frame limiter. If that is not available (or not well implemented), RTSS is usually the best option.

 

[SapientWolf]

It sounds like variable refresh rate support is a must have for every PC game, regardless of the framerate limit. 144+hz offers a massive reduction in input lag compared to 60hz.

 

[Hellshingu]

So does having RTSS set to 141 global and a game like Overwatch at 141 using their frame rate limiter overwrite the option?

I have my RTSS set to 141 on global just so I don't have to configure each game separately but wouldn't mind doing so if I needed to.

 

[Canklestank]

Welp, now my 60 Hz monitors feel super inadequate. Maybe someday I'll have G-Sync.

Bookmarked for the future, just in case. Good read, nonetheless.

 

[The Hamburglar]

TL;DR people:

Gsync On + Vsync On + Framelimit -3fps below your max

Ex: 144hz monitor, you should cap your framerate at 141

This gives you the stutter free, no tearing and no input lag experience.

What if you have a 144 monitor but can't hit 144 fps in games due to the game you're playing. Does capping at 140 still help?

 

[Durante]

What if you have a 144 monitor but can't hit 144 fps in games due to the game you're playing. Does capping at 140 still help?

If your framerate is below the maximum refresh rate with G-sync you are already as close to the ideal in terms of input lag as you can get at that framerate.