Page 1 of 2

[Troubleshoot] Why Vsync had better latency than frame cap with Freesync? [FIXED]

Posted: 27 Jul 2023, 03:24
by Stizzie
Contrary to the well explained and articulated GSYNC 101 guide, which state that the configuration of Vsync, Gsync and a few frames cap under max refresh rate of the display is the optimal terms of latency. I find the same rule does not apply well with Freesync.

In my experience, Freesync works well with Vsync. However, when a frame cap is introduced, whether FRC or RTSS, input latency becomes much worst. I'm not talking about those elaborate measures where milliseconds are taken into account, I'm talking the real feels when I move my cursor or movement input. Now I'm not one of those people who sensitive to the milliseconds of input delay, as long as games are fluid and responsive, I'm good. Still the difference between having a frame cap and none is apparent enough for me to not notice there's something wrong with my input.


The Gsync 101 guide never mentioned that it is applicable to Freesync, though I've came across few people saying that is it. AMD also don't mention anything about having a frame cap in their recommended settings for Freesync. So I'd like to know what is your experience on this.

[Troubleshoot] Why Vsync had better latency than frame cap with Freesync? [FIXED]

Posted: 27 Jul 2023, 08:23
by jorimt
Stizzie wrote:
27 Jul 2023, 03:24
In my experience, Freesync works well with Vsync. However, when a frame cap is introduced, whether FRC or RTSS, input latency becomes much worst. I'm not talking about those elaborate measures where milliseconds are taken into account, I'm talking the real feels when I move my cursor or movement input.
FreeSync + V-SYNC + FPS above the refresh rate = standalone V-SYNC behavior, which equals anywhere from 2 to 4 frames of extra latency vs. the same scenario with the framerate within the refresh rate. Whatever you're experiencing must be specific to your configuration and/or monitor model.
Stizzie wrote:
27 Jul 2023, 03:24
The Gsync 101 guide never mentioned that it is applicable to Freesync, though I've came across few people saying that is it.
My base optimal VRR settings (VRR + V-SYNC + minimum -3 FPS limit) apply to AMD FreeSync as well.

All forms of VRR are fundamentally the same; they only function whenever the framerate is within the refresh rate, and without the V-SYNC option enabled, tearing can still occur.

[Troubleshoot] Why Vsync had better latency than frame cap with Freesync? [FIXED]

Posted: 27 Jul 2023, 11:56
by Stizzie
jorimt wrote:
27 Jul 2023, 08:23

FreeSync + V-SYNC + FPS above the refresh rate = standalone V-SYNC behavior, which equals anywhere from 2 to 4 frames of extra latency vs. the same scenario with the framerate within the refresh rate. Whatever you're experiencing must be specific to your configuration and/or monitor model.


My base optimal VRR settings (VRR + V-SYNC + minimum -3 FPS limit) apply to AMD FreeSync as well.

All forms of VRR are fundamentally the same; they only function whenever the framerate is within the refresh rate, and without the V-SYNC option enabled, tearing can still occur.
Thanks for the reply! Didn't expect one from the author himself.

So if I understand correctly, a framerate equal to refresh rate (Vsync ON) is considered out of VRR range ? Because Vsync ON essentially cap your framerate at refresh rate, not above.

I remember reading a post about Vsync overshoots which said that Vsync occasionally let framerate go above refresh rate thus make VRR revert to ( or as stated in your guide, try to mimic ) Vsync behaviors. Well considering that would only intermittently happen I'm content to compromise. Why ?

Because limiting framerate could be finicky, especially to those game without an in-game limiter. In fact, my experience with NFS Payback is what I refer to in the first post. The game just doesn't like to be capped. If limit is applied it feels very sluggish, which I might misinterpreted as an increase in input delay.

Re: I find Vsync has much better latency than frame cap when use with Freesync

Posted: 27 Jul 2023, 18:36
by jorimt
Stizzie wrote:
27 Jul 2023, 11:56
So if I understand correctly, a framerate equal to refresh rate (Vsync ON) is considered out of VRR range ?
Yes.
Stizzie wrote:
27 Jul 2023, 11:56
I remember reading a post about Vsync overshoots which said that Vsync occasionally let framerate go above refresh rate thus make VRR revert to ( or as stated in your guide, try to mimic ) Vsync behaviors. Well considering that would only intermittently happen I'm content to compromise. Why ?

Because limiting framerate could be finicky, especially to those game without an in-game limiter. In fact, my experience with NFS Payback is what I refer to in the first post. The game just doesn't like to be capped. If limit is applied it feels very sluggish, which I might misinterpreted as an increase in input delay.
Up to you, but again, if you're using VRR + V-SYNC and the framerate is maintained "at" the refresh rate (V-SYNC still let's frametime "exceed" average framerate periodically, causing a build-up in the buffer, hence extra latency), it reverts to standalone V-SYNC behavior.

[Troubleshoot] Why Vsync had better latency than frame cap with Freesync? [FIXED]

Posted: 27 Jul 2023, 18:42
by Chief Blur Buster
FYI, what Hz are you? Classically the recommendation was 3fps below. The new recommendation is usually ~3% below, e.g. 360Hz should be capped about 10-15fps below.

There is some anecdotal evidence that for less performant VRR implementations ("G-SYNC Compatible" and "FreeSync") the error margin may need to be bigger. NVIDIA Reflex has a behavior like this; where it uses a bigger safety margin than 3fps.

Re: I find Vsync has much better latency than frame cap when use with Freesync

Posted: 28 Jul 2023, 01:15
by Stizzie
Chief Blur Buster wrote:
27 Jul 2023, 18:42
FYI, what Hz are you? Classically the recommendation was 3fps below. The new recommendation is usually ~3% below, e.g. 360Hz should be capped about 10-15fps below.

There is some anecdotal evidence that for less performant VRR implementations ("G-SYNC Compatible" and "FreeSync") the error margin may need to be bigger. NVIDIA Reflex has a behavior like this; where it uses a bigger safety margin than 3fps.
hi Chief ! I'm running a 75Hz LG. In that case I may as well cap at 70/75Hz for absolute margin safety then.

While you guys at it, mind if I have a extra question:

I know that in tradional Vsync if a new frame is not ready after the VBI then another scan out of the previous frame will take place until the new frame is ready while Gsync would hold on to the VBI until a new frame (please correct me if I'm wrong).

So how the VSYNC+GSYNC+Frame limit config's behave on this. Does it behave like VSync or Gsync ?

[Troubleshoot] Why Vsync had better latency than frame cap with Freesync? [FIXED]

Posted: 28 Jul 2023, 08:18
by jorimt
Stizzie wrote:
28 Jul 2023, 01:15
In that case I may as well cap at 70/75Hz for absolute margin safety then.
While the Chief is generally correct about a lower limit, especially for higher than 240Hz refresh rates (which is why my recommendation is a minimum -3 FPS limit, and you're free to limit the FPS as low as you like), -5 FPS at 75Hz should be more than enough to keep the framerate within the G-SYNC range; my original tests, for instance, were conducted at 240Hz -2 FPS in the article (most of which were with an in-game limiter).

What is your monitor model? Because 75Hz is a relatively low G-SYNC range by today's standards, and not all monitor VRR implementations are created equal, so it's possible the issues you've been experiencing are down to limitations with your model.
Stizzie wrote:
28 Jul 2023, 01:15
I know that in tradional Vsync if a new frame is not ready after the VBI then another scan out of the previous frame will take place until the new frame is ready while Gsync would hold on to the VBI until a new frame (please correct me if I'm wrong).

So how the VSYNC+GSYNC+Frame limit config's behave on this. Does it behave like VSync or Gsync ?
Both G-SYNC and V-SYNC will repeat the previous frame if/when a new frame isn't ready. In fact, no sync will as well. The only difference between the three is how they handle tearing.

I think what you're referring to about G-SYNC is what I explain in entry #2 of my Closing FAQ:
https://blurbusters.com/gsync/gsync101- ... ttings/15/

As for framerate limiting, so long as said limit keeps the framerate well enough within the refresh rate, it doesn't directly affect whether G-SYNC behaves like V-SYNC with the VBI or not as much as it ensures the frametime of the framerate remains within the frametime of the physical refresh rate's scanout time so G-SYNC doesn't revert to standalone no-sync (G-SYNC on + V-SYNC off) or V-SYNC (G-SYNC on + V-SYNC on) behavior.

[Troubleshoot] Why Vsync had better latency than frame cap with Freesync? [FIXED]

Posted: 01 Aug 2023, 20:38
by Chief Blur Buster
Jorim is correct too.

Stizzie, you only have a 75Hz VRR monitor?
In that case... 75Hz doesn't need a big cap-below.

But to muddy the waters further -- there's even exceptions to the 3fps-below rule. You can even get by with 1-2fps cap-below quite easily. The 144Hz worked well with a 3fps-below cap, so half refresh rate (75Hz) could easily get by with a 2fps-below cap.

If you've got very accurate frame pacing (e.g. emulator 60fps), you can even get by with 0.5fps differential. There are 4K60 display owners (40-60Hz VRR range) who can overclock to 60.5Hz and then cap to 60fps, to get low-latency VRR in an NTSC 59.94fps or 60.00fps emulator.

Capping margin is more time based (e.g. giving the drivers/monitor enough breathing room to avoid switching to laggy VSYNC ON) -- so lower Hz (than the refresh rates Jorim tested at) can even get by with less than 3fps-below.
- Refresh rate range you're dealing with
- Game framepacing accuracy
- Driver framepacing accuracy
- Frame capping accuracy (microsecond-accurate cappers like RTSS can get by with tighter margins than in-game framerate cpping)
- Power management (Performance Mode can get away with tighter margins)
- Monitor-side LFC (GSYNC Native) versus Driver-side LFC (FreeSync/VESA AdaptiveSync/G-SYNC "Compatible"/Generic VRR)

The more inaccurate/jittery everything is, the more safety margin you need to prevent those lovely VRR frames from turning into laggy VSYNC ON frames or teary VSYNC OFF frames (and its assocated annoying sudden latency-change effects). Current esports players who wish to use VRR in esports without capping, can just simply purchase more VRR range than their framerate range (e.g. say, 30Hz-540Hz for a new 540Hz G-SYNC native VRR displays), and play comfortably completely inside their VRR range without the capping compromise. But here, we're dealing with optimizing VRR on an existing low-Hz monitor...

However, as a general rule of thumb, 3fps is a good minimum boilerplate recommendation -- unless you know what you're doing.

Then you can tighten it even further on a case-by-case basis (e.g. very accurate framepaced emulators that rarely output "slow" 1/60.5sec frames to violate a tight cap).

The good thing is you've got a 75Hz max Hz, so that will keep all your 60fps games pretty low latency, since 60fps content, by definition, is self-capping (console ports, emulators, etc).

For your situation, a 72fps cap is a pretty comfy cushion, possibly a bit more than needed, and nicely accomodates in-game 24fps movie material (24fps x 3 = 72Hz).

[Troubleshoot] Why Vsync had better latency than frame cap with Freesync? [FIXED]

Posted: 06 Aug 2023, 01:01
by Stizzie
the water is clear now Chief !

I want to make a correction for someone maybe come across this thread because now the title is rendered wrong information (it was at first place, my bad). Vsync is worst latency wise. But capping frame also poses frame pacing isssue for some games. I think it's a matter of picking your poison. For me personally I'm willing to use Vsync+Gsync with no cap for those games. Guess I'm more susceptive to percieved smoothness than input latency.

In instances where 60 fps is enough. Can I change the refresh rate to 60hz, essentially manipulate the Vsync cap to 60 ? Does VRR work with every refresh rates in its range or it has to be the native max refresh rate ?

[Troubleshoot] Why Vsync had better latency than frame cap with Freesync? [FIXED]

Posted: 06 Aug 2023, 09:55
by jorimt
Stizzie wrote:
06 Aug 2023, 01:01
I want to make a correction for someone maybe come across this thread because now the title is rendered wrong information (it was at first place, my bad). Vsync is worst latency wise. But capping frame also poses frame pacing isssue for some games. I think it's a matter of picking your poison. For me personally I'm willing to use Vsync+Gsync with no cap for those games. Guess I'm more susceptive to percieved smoothness than input latency.
Which is why I asked what your model was, because not all 75Hz FreeSync models support LFC and/or a large enough LFC range, which means some of them can drop out of VRR behavior both above and "below" a certain refresh rate range, potentially causing more stutter during any frametime fluctuations vs. standalone V-SYNC, in-which case non-VRR is indeed probably smoother for you in most cases so long as the framerate can be sustained at/above the refresh rate.
Stizzie wrote:
06 Aug 2023, 01:01
In instances where 60 fps is enough. Can I change the refresh rate to 60hz, essentially manipulate the Vsync cap to 60 ? Does VRR work with every refresh rates in its range or it has to be the native max refresh rate ?
VRR should work at any physical refresh rate (that the given monitor supports), but lowering the physical refresh rate itself will just slow frame delivery, regardless of achievable framerate with VRR due to the reduction in scanout speed.