Variable refresh and input lag?

[Gdourado]

Hello, how are you?

I was reading an article about g-sync and freesync.

I read that with freesync, the frame rate can never go above the refresh rate of the monitor.

And that with freesync, the user has the option to either use v-sync when the frame rate is higher than the refresh or not.

If no v-sync is used, then tearing can occur even with freesync.

My question...

From what I read, v-sync can resolve the screen tearing but as a negative aspect, it adds input lag, and that is why before the variable refresh rates, many people game with v-sync off.

Since g-sync caps the fps to avoid tearing and freesync needs v-sync to also cap the fps to avoid tearing, that g-sync and freesync also add input lag?

Is this corrects, being the only benefit of the technology to avoid stutter when frames drop from the maximum refresh, but still introduce input lag?

I am a bit confused...

Cheers!

[minibois]

Ok, ok. You're throwing quite a few technologies in a mix now..

Let me clear some stuff up:

 

Vsync:

Something that limits your fps to your monitor's refresh rate (so if 60Hz monitor, 60fps max, 144Hz monitor means 144fps, etc.)

 

Freesync/Gsync:

Something that takes the fps rate you're getting and making it so your monitor has the refresh rate of whatever you have.

 

Vsync should make it so your fps can't go above your monitor's refresh rate. Not sure if Gsync or freesync do that too.

I would suggest watching the TechQuickie videos on these subjects

[Gdourado]

Took this:

One other feature that differentiates FreeSync from G-SYNC is how things are handled when the frame rate is outside of the dynamic refresh range. With G-SYNC enabled, the system will behave as though VSYNC is enabled when frame rates are either above or below the dynamic range; NVIDIA's goal was to have no tearing, ever. That means if you drop below 30FPS, you can get the stutter associated with VSYNC while going above 60Hz/144Hz (depending on the display) is not possible – the frame rate is capped. Admittedly, neither situation is a huge problem, but AMD provides an alternative with FreeSync.

Instead of always behaving as though VSYNC is on, FreeSync can revert to either VSYNC off or VSYNC on behavior if your frame rates are too high/low. With VSYNC off, you could still get image tearing but at higher frame rates there would be a reduction in input latency. Again, this isn't necessarily a big flaw with G-SYNC – and I’d assume NVIDIA could probably rework the drivers to change the behavior if needed – but having choice is never a bad thing.

[skullbringer]

Your assumption would be correct if adaptive sync would change the hz of your monitor, which it does.

With regular v-sync you would need to wait about 16ms (60Hz panel) for a frame to be dispayed on the monitor.

 

But with variable refresh rate technologies, the monitor updates as soon as the gpu is finished drawing the frame, so in theory you have no signal input lag.

[Sakkura]

But with variable refresh rate technologies, the monitor updates as soon as the gpu is finished drawing the frame, so in theory you have no signal input lag.

 

But he's talking about the situation where you're going above the maximum refresh rate of the Freesync monitor. So eg. running 90FPS with a Freesync monitor that supports 30-60Hz, or running 180FPS on a Freesync monitor that supports 48-144Hz. In that case you will get the same input lag as you would with Vsync, assuming you tell the driver to use Vsync when you go outside what the monitor can handle dynamically. It's just that it'll be a pretty small input lag especially in the 144Hz case.

[skullbringer]

But he's talking about the situation where you're going above the maximum refresh rate of the Freesync monitor. So eg. running 90FPS with a Freesync monitor that supports 30-60Hz, or running 180FPS on a Freesync monitor that supports 48-144Hz. In that case you will get the same input lag as you would with Vsync, assuming you tell the driver to use Vsync when you go outside what the monitor can handle dynamically. It's just that it'll be a pretty small input lag especially in the 144Hz case.

but thats the whole point, there is no static refresh rate, you dont need to wait for anything.

[CostcoSamples]

My question...

From what I read, v-sync can resolve the screen tearing but as a negative aspect, it adds input lag, and that is why before the variable refresh rates, many people game with v-sync off.

Since g-sync caps the fps to avoid tearing and freesync needs v-sync to also cap the fps to avoid tearing, that g-sync and freesync also add input lag?

Is this corrects, being the only benefit of the technology to avoid stutter when frames drop from the maximum refresh, but still introduce input lag?

I am a bit confused...

Cheers!

 

Gsync is a circuit board inside the monitor that allows your GPU to control the monitor's refresh.  That way, each time a frame is rendered, it is immediately displayed on screen.  Vsync is a software solution that forces your GPU to match the fixed refresh rate of the monitor.  So, with Gsync, the GPU will simply output frames as fast as it can, and the monitor follows.  Think of it like a dance.  The man is the GPU, and he is not a very good dancer.  The woman is the monitor, and she has perfect rythm and movement.  If the woman leads, the man will always be fumbling to catch up (input lag).  If the man leads, he may go too fast or too slow at times, but the woman can easily match his steps.

 

So no, there shouldn't be any input lag from Gsync.  The situation where fps goes above or below the Gsync window (say 30 fps minimum and 60 fps maximum, for a 60hz panel) is where the monitor must take over to some degree.  The monitor cannot operate below 30hz or above 60hz, so in that situation gsync will be disabled until fps returns to the gsync operating parameters.  I think the panel returns to its default refresh rate in that case.   When the GPU is operating outside gsync parameters, in the low fps case vsync takes over, which would indeed add input lag.  In the case of very high fps, it should simply cap GPU fps output at the max refresh rate which would not require gsync to disable.

[Sakkura]

but thats the whole point, there is no static refresh rate, you dont need to wait for anything.

 

Yes there is, when you go outside the range supported by Freesync (or G-sync).

[Glenwing]

Your assumption would be correct if adaptive sync would change the hz of your monitor, which it does.

With regular v-sync you would need to wait about 16ms (60Hz panel) for a frame to be dispayed on the monitor.

But with variable refresh rate technologies, the monitor updates as soon as the gpu is finished drawing the frame, so in theory you have no signal input lag.

Well, it's 16ms from the last frame, not 16ms from your input. If your monitor has 2ms input lag and you click 3ms before the next refresh, you'll only have to wait 3ms for it to show. Dispays have an actual signal processing delay independent of refresh cycle. Your measured input lag would be anywhere between 2-18ms ("real" input lag plus 0-16ms) and you need to take many measurements to properly determine the best case input lag. The actual input lag varies each time between the best case plus the cycle length (16ms) so with variable refresh, the cycle length factor is removed. But you're still left with some amount of best-case input lag, not 0 lag.

[Glenwing]

but thats the whole point, there is no static refresh rate, you dont need to wait for anything.

The monitor does have a limit to how fast it can refresh. If your framerate is above that limit, the panel won't refresh as soon as a new frame is received anymore. That's the situation we're talking about.

[skullbringer]

Well, it's 16ms from the last frame, not 16ms from your input. If your monitor has 2ms input lag and you click 3ms before the next refresh, you'll only have to wait 3ms for it to show. Dispays have an actual signal processing delay independent of refresh cycle. Your measured input lag would be anywhere between 2-18ms ("real" input lag plus 0-16ms) and you need to take many measurements to properly determine the best case input lag. The actual input lag varies each time between the best case plus the cycle length (16ms) so with variable refresh, the cycle length factor is removed. But you're still left with some amount of best-case input lag, not 0 lag.

Yeah, thats what I meant to say. The input lag is not the vsync lag, but physics.

[Guest Strangerbob]

Yeah, thats what I meant to say. The input lag is not the vsync lag, but physics.

 

Good thing that on a 144hz monitor the Vsync input lag is very hard to notice.

[skullbringer]

Good thing that on a 144hz monitor the Vsync input lag is very hard to notice.

in most cases the os and game create the most input lag, the hardware side is basically unnoticeable. 

[Guest Strangerbob]

in most cases the os and game create the most input lag, the hardware side is basically unnoticeable. 

U should try 60/60 vsync and then 144/144 vsync on the same hardware. Let me know what you think afterwards.

[skullbringer]

U should try 60/60 vsync and then 144/144 vsync on the same hardware. Let me know what you think afterwards.

of course, 6.9ms are less that 16.6ms of input lag, but thats not the point here.