Input Lag G-Sync vs V-Sync vs No-Sync

[Irlwizard]

Q1) Which creates the most input lag on the same monitor, has anyone tested the different options on a 1080p G-Sync monitor?

a) G-Sync

b) V-Sync + Tripple Buffering

c) V-Sync + Buffering Disabled

d) V-Sync OFF

Q2) If you have a 144Hz monitor or even 120Hz monitor and the graphics horsepower to constantly produce over 150FPS in CS:GO. Are any of the sync options even necessary to minimize stutter?

Q3) What is the difference between V-Sync and setting your game to run fixed at e.g. 60FPS if it is a new title that you can't run at your monitors max Hz rate? Or perhaps you would like to set a cap to game without the fans going nuts during a RPG so you can chilll at a silent 60FPS. Due to the existence of FPS console capping exists, V-Sync seems pointless to me.

[Sparky]

There's a whole section for input lag.



G-sync latency is equivalent to v-sync off when running below the maximum refresh rate, and is equivalent to double buffered v-sync when bottlenecked by the display. You can get around the latter with a framerate cap(preferably in-game).

V-sync when bottlenecked by the display adds a lot of input lag. Triple buffering adds 1 extra frame when display bottlenecked.
V-sync when not bottlenecked by the display adds an average of 1/2 frame of latency. Triple buffering can get a slightly better average here, due to framerate not dropping to half refresh rate, but you still get stutter from missing frames.

Q2: depends on the person, at high framerates you'd probably prefer v-sync off and ULMB on, though you still get tearing, each tear is just on screen for less time.

[Irlwizard]

There's a whole section for input lag.

Thank you for your response and sorry if I posted in the wrong forum section.

V-sync when bottlenecked by the display adds a lot of input lag.

Q1) So you are saying if the screen is 60Hz and your GPU is producing 84 FPS or whatever, the input lag kicks in? Or are you saying that input lag kicks in if the FPS is below the monitors refresh rate? When you say "bottleneck" I don't know if you mean when the FPS is ABOVE or BELOW 60Hz.

Triple buffering adds 1 extra frame when display bottlenecked.

V-sync when not bottlenecked by the display adds an average of 1/2 frame of latency.

Triple buffering can get a slightly better average here, due to framerate not dropping to half refresh rate, but you still get stutter from missing frames.

According to this old and probably outdated anandtech Vsync topic below. Tripple buffering should always be on regardless if Vsync is on or not and according to him it 'eliminates stutter'. Sounds like Gsync to me or a very close alternative, but you say it comes with input lag? http://www.anandtech.com/show/2794/4

[Sparky]

There's a whole section for input lag.

Thank you for your response and sorry if I posted in the wrong forum section.

V-sync when bottlenecked by the display adds a lot of input lag.

Q1) So you are saying if the screen is 60Hz and your GPU is producing 84 FPS or whatever, the input lag kicks in? Or are you saying that input lag kicks in if the FPS is below the monitors refresh rate? When you say "bottleneck" I don't know if you mean when the FPS is ABOVE or BELOW 60Hz.

Triple buffering adds 1 extra frame when display bottlenecked.

V-sync when not bottlenecked by the display adds an average of 1/2 frame of latency.

Triple buffering can get a slightly better average here, due to framerate not dropping to half refresh rate, but you still get stutter from missing frames.

According to this old and probably outdated anandtech Vsync topic below. Tripple buffering should always be on regardless if Vsync is on or not and according to him it 'eliminates stutter'. Sounds like Gsync to me or a very close alternative, but you say it comes with input lag? http://www.anandtech.com/show/2794/4

The big issue with that article is that the triple buffering implementations don't drop frames like that article implies. If it did work like that it would have lower latency, but you'd also be rendering a lot of frames that never get displayed, which means more heat, noise, and power consumption.

[Irlwizard]

Well if the article is wrong, not even Anandtech knows the difference between Vsync with no triple buffering and Vsync with triple buffering on. Then how is the average Joe supposed to understand it? What the hell is the difference? What is a "half frame of latency" can't a number be put on it? Is that a 0.1ms delay or what? Very negligible considering human reaction time.

[flood]

anandtech is a website for casuals :DDD

[Irlwizard]

anandtech is a website for casuals :DDD

Yeah I hate it the guy thinks he is an expert and spreads misleading information this was also the case when I was trying to learn about load line calibration back in the day on X58 systems when I was a extreme overclocker. I never liked anandtech but it seems to have a high rep in the community, god knows why... Blurbuster has been a lot better experience, monitors just seem more complicated than PC systems in general.

[Sparky]

Well if the article is wrong, not even Anandtech knows the difference between Vsync with no triple buffering and Vsync with triple buffering on. Then how is the average Joe supposed to understand it? What the hell is the difference?

The difference, in normal default implementations, is that double buffering drops you down to half your refresh rate when your framerate would otherwise be between half your refresh rate and your refresh rate. triple buffering means you can half framerates in between, but it has an extra frame of latency over double buffered vsync when your framerate would otherwise be above your refresh rate.

With v-sync off there are two framebuffers, one is being scanned out to the display, and the other is having the next frame drawn in it. When the frame is done rendering, those buffers swap places(called a buffer flip). If you're halfway done sending the frame to the display, this causes a tear, because the top half of the screen was the old frame, and the bottom half is the new frame. When you enable v-sync, the buffer flip is delayed until the previous frame is done getting sent to the display. This means there's nowhere for the GPU to start working on the next frame, so your frame time jumps up to the next multiple of refresh rate. Ideally that's framerate = refresh rate.

With triple buffering, there's a third buffer, so the GPU can keep working instead of waiting on the next buffer flip, but in the current implementations, all the frames are displayed, so at max framerate triple buffering will just fill up both buffers, so you have 1 extra frame of latency over double buffering.

With g-sync, the GPU behaves a lot like v-sync off, and the display just displays the frame whenever it's done rendering.
With freesync, well, we don't have as much information on exactly how freesync works, but it should be something like g-sync with triple buffering instead of hardware inside the display. If the frame is finished within the VRR window it gets displayed on time, otherwise the previous frame gets sent again. At least, that's what I think AMD's current drivers are supposed to do, but there's a lot of room for improvement on what happens below the VRR window.

What is a "half frame of latency" can't a number be put on it? Is that a 0.1ms delay or what? Very negligible considering human reaction time.

1 frame of latency is 1/framerate. So at 60fps, one frame of latency would be 16.7ms, at 100fps one frame would be 10ms, etc.