Blur Buster's G-SYNC 101 Series Discussion

[Chief Blur Buster]

144 Hz monitors can be switched to many refresh rates lower, e.g. 60Hz, 120Hz, 144Hz, etc.

Also, consider GSYNC or FreeSync.

You can get a variable refresh rate display which means you can play any framerate without stutter. It looks like a permanently-smooth-effect-at-any-framerate where changing framerates has no stutter. If you have never seen a variable refresh rate, see the simulation at www.testufo.com/vrr (make sure you quit all applications & browser windows), launch a stutterless web browser (to avoid stutters from interfering with the demo) and try the Slow Ramp, Fast Ramp, Random, Struggle At Max. You should ideally consider VRR in your next 144 Hz display (FreeSync or G-SYNC).

The only other option I see is going up to 240hz which I definitely wouldn't need

Ten years ago, nobody needed 144 Hz ... In a few years, 240 Hz will have no premium over 144 Hz. 240 Hz has half the motion blur of 120 Hz, and 120 Hz has half the motion blur of 60 Hz. 240 Hz would have one-quarter the motion blur of 60 Hz. Twenty years ago, nobody needed 1080p, but today the only HDTVs you can easily buy in Best Buy are all 4K now, since they've cleared out the majority of 1080p already (except for small TVs like 32").

That said, 144 Hz is definitely the mainstream gaming monitor resolution now. Anybody buying a desktop PC gaming monitor *today* should buy triple digits.

[KKNDT]

This is from the article

"With V-SYNC OFF at -2 FPS below the refresh rate, for instance (the scenario used to compare V-SYNC OFF directly against G-SYNC in this article), the tearline will continuously roll upward, which means, when measured by first on-screen reactions, its advantage over G-SYNC can be anywhere from 0 to 1/2 frame, depending on the ever-fluctuating position of the tearline between samples. "

Why does the tearline move upward? I think the tearline should move downward when FPS<HZ.

How to understand that V-SYNC OFF's lag advantage over G-SYNC is 1/2 frame tops?

[RealNC]

The article is wrong here

[Chief Blur Buster]

Why does the tearline move upward? I think the tearline should move downward when FPS<HZ.

Good catch. One-word correction applied. Even myself made that mistake. (It's been a while, but I suspect I was the one who provided that very specific quote to Jorim about scrolling tearlines)

Yes, for framerate-near-Hz situations during VSYNC OFF, the tearline scrolls downwards for bigger frametimes (lower framerates) and upwards for smaller frametimes (bigger framerates). Easy to confuse bigger/smaller -- sine frametimes are inverse to framerates.

e.g.
The tearline scrolls downwards in a 1-wraparound-every-2-seconds cycle at 143.5fps @ 144Hz (slow scrolling tearline)
The tearline scrolls downwards in a 1-wraparound-every-1-seconds cycle at 143fps @ 144Hz
The tearline scrolls downwards in a 2-wraparound-every-1-seconds cycle at 142fps @ 144Hz
The tearline scrolls downwards in a 3-wraparound-every-1-seconds cycle at 141fps @ 144Hz (faster scrolling tearline)

For framerate < refreshrate, scrolling-tearline is downwards, with wraparounds per second being (refreshrate - framerate).
For framerate > refreshrate, scrolling-tearline is upwards, with wraparounds per second being (framerate - refreshrate).

Assuming frametimes stay very consistent, of course.

The more varying the frametimes are, the more the tearline will jitter/vibrate. The tearline position represents 1 of horizontal scanrate, so if the display signal horizontal scanrate (as seen in CRU) is 135 KHz (135,000 scanlines per second), a frametime variance being delayed 1/135,000th of a second later than average frametime, means a 1 pixel shift downwards in the tearline from the average-frametime predicted location of the tearline.

Needless to say, frametimes are rarely that super-consistent, so a slow-scrolling tearline will usually vibrate quite a bit. Given sufficient frametime variations, or sufficient difference between framerate-and-refreshrate, the tearline just looks random all over the screen.

It's only when framerate-near-refreshrate (or near a harmonic frequency, e.g. half framerate or double framerate) with fairly consistent frametimes, that you get a noticeably visible scrolling-tearling effect during horizontal motion (strafing, turning, horizontal scrolling, etc).

[Chief Blur Buster]

How to understand that V-SYNC OFF's lag advantage over G-SYNC is 1/2 frame tops?

And on this...

GSYNC has a mandatory scanout lag for center (1/2 refresh cycle) and bottom (full refresh cycle). At 240Hz, that's pretty low scanout lag (scanout lag modifier to top-edge absolute lag would be would be +2ms center, +4ms bottom, relative to the top edge -- 1/240sec is 4.166666ms). Average all of that, for all pxiels combined, and you come with +2ms average lag modifier.

Contrast to VSYNC OFF which has zero scanout lag for the pixels immediately below the tearline --no matter where the tearline is. That said, frame slices are lag gradients themselves, so to get the maximal benefits, you'd need infinite framerate to get the full half-refresh-cycle lag savings.

That said, 2000 frames per second VSYNC OFF would have only a 1/2000sec = 0.5ms lag graident for frame slices (0ms for top edge of frame slice, 0.25ms for center of frameslice, and 0.5ms for bottom of frame slice -- center of frame slice is equal to average lag too -- so 0.25ms scanout input lag for that frameslice). So in this case, 2000fps VSYNC OFF has a +0.25ms scanout lag modifier (to absolute lag), and 240fps GSYNC has a +2ms scanout lag modifier (to absolute lag).

Average scanout-induced lag modifier of GSYNC refresh cycle equals an addition of half a refresh cycle.
Average scanout-induced lag modifier of VSYNC OFF frame slice equals an addition of half a frame time.

That said, position of lag modifiers can matter if you've got a fixed element of lag importance (e.g. a crosshairs in the screen centre -- which is super duper important, you know!). For VSYNC OFF, the scanout lag addition can vary from 0...frametime for any pixel anywhere on the screen, while for GSYNC, it's always consistent (+2ms for center, +4ms for bottom) based on the scanout velocity. Because it's a permanent scanout lag for crosshairs location during VRR, it feels like a +2ms absolute lag in screen centre. Only the most competitive elite players will find that an issue.

(I say lag modifier because lag is caused by many factors, scanout is simply one of them, and shifts the averaged input lag)

Needless to say, CS:GO does not run at such high framerates.

Also, 240 Hz GSYNC has more advantages than VSYNC OFF in many games that often run at framerates below refresh rates -- like PUBG which fluctuates framerates all over the place -- you can score more hits with 240 Hz GSYNC than VSYNC OFF when you're playing PUBG instead of CS:GO because of the way PUBG stuttering often interferes with your aiming. Whereas CS:GO often runs at such high framerates, aiming is often more locked-feeling for VSYNC OFF. Everyone has their preferences in their favorite sync modes but there are certain games where one likes using the right-tool-for-right-job.

[RealNC]

vsync off at the same FPS cap as gsync having up to 1/2 frame less lag is difficult to explain though. I don't think this is the case at all.

If you cap to the same FPS (2 below Hz), vsync off vs gsync should have identical latency, with the only difference being gsync overhead (if there is any.)

[Chief Blur Buster]

vsync off at the same FPS cap as gsync having up to 1/2 frame less lag is difficult to explain though. I don't think this is the case at all.

It has more to do with MIN/MAX/AVG in many ways.

For VSYNC OFF, depending on position of tearlines, the lag can vary [0...frametime] relative to the tearline positions, while for GSYNC, the lag vary [0....refreshtime] from top to bottom edge.

So for screen centre, same-framerate-as-refreshrate situation, crosshairs VSYNC OFF is [varying 0...frametime] while crosshairs GSYNC is [more consistent 1/2 refresh cycle] when measuring lag-at-crosshairs rather than first-anytime-reaction. The average lag can be similar the same, but you get random-chance zero-lag with VSYNC OFF screen centre. Sometimes the chance zeros are important to some paid eSports players....

So that's why if you want to play GSYNC in eSports (with paid money prizes), you want to minimize the scanout lag and use the fastest scanout velocity available since GSYNC is a top-to-bottom finish rather than "interrupt-scan-at-the-spot-of-tearline" (VSYNC OFF). So 240 Hz GSYNC minimizes the scanout lag penalty.

[noobysnackz]

Sorry if this has been answered. I have read the full article and I still have a few questions. With optimal settings are we talking 20ms compared to with everything shut off? Or is for example 15ms delay no matter what and gsync adds only 5ms extra? All this is hard for me to understand I wish there was a conclusion on what games you think its worth using on and games you think you shouldn't use it on.
I play mostly PUBG,OW,and Fortnite. Does having Gsync on but not capping frames add delay or is it basically doing nothing since i am not capping frames or using vsync in nvcp thats how i have been running it. I have the ASUS ROG Swift PG248Q I play at 144hz not at 180hz and even though i tried the optimal setting im just not sure any delay is worth it or not yet but i do admit it looks so smooth.

[RealNC]

So for screen centre, same-framerate-as-refreshrate situation, crosshairs VSYNC OFF is [varying 0...frametime] while crosshairs GSYNC is [more consistent 1/2 refresh cycle] when measuring lag-at-crosshairs rather than first-anytime-reaction. The average lag can be similar the same, but you get random-chance zero-lag with VSYNC OFF screen centre. Sometimes the chance zeros are important to some paid eSports players....

In this case you can use gsync with vsync off. These "rogue" frames should then tear.

Although it doesn't make sense to consider "rogue" frames as an actually important factor in latency. Because right now, these rogue frames are the result of imperfect frame limiters. If you say that imperfect frame limiters have a latency advantage, then you need to consider that the reason for that is because they sometimes allow the frame rate to be exceeded or not reached every so often. Well, duh! If you exceed or don't reach the target frame rate, then that just means the frame cap didn't get applied correctly. If you consider that an advantage, then don't frame cap at all!

[Chief Blur Buster]

In this case you can use gsync with vsync off. These "rogue" frames should then tear.

Decision whether to use GSYNC+VSYNC OFF versus just playing pure VSYNC OFF -- depends on how sensitive you are to lag mechanics changes.

For the rareified professional elite players -- gaming for real-money prize pots -- the sudden change in lag mechanics can throw off "what you're used to" in aiming, where below VRR max, scanout lag modifier at crosshairs is a constant [+half top-Hz refresh cycle] addition to absolute lag, and then above VRR in VSYNC OFF, scanout lag modifier at crosshairs suddenly instead becomes a [+0...+frametime] lag lottery effect. A lag-mechanics change.

Simplified mathematical example -- 8000 pixels horizontal panning during turns -- flick-180-turn-straight-into-aim-behind-you -- 1ms error equals 8 pixels misaim -- so milliseconds of lag-mechanics changes will be noticed in "I feel like I am not getting frags nearly as frequently as I normally do" rareified esports players where single milliseconds often matters more to such individuals. At 144Hz VRR adding a sudden 6.9ms lottery factor may mean a full ~55 pixels range of misaim (overshoot/undershoot). They may not feel the lag directly but seasoned eSports player -- they may immediately notice "eh, I seem to be missing my targets twice as often now" or "eh, I seem to be suddenly overshooting my flicks more often." -- even single milliseconds can do that to such well-attuned individuals. Blur Busters does not dismiss the unanticipated surprising importance of milliseconds to such individuals.

Alas, all of these are not necessarily easy to scientifically measure, though I'm certainly willing to commission research and studies -- like spacediver's guest article (Input Lag And The Limits Of Human Reflex) -- inquire within (squad[at]blurbusters.com) if any researchers are reading this thread.

Super-trained people super-sensitive to lag-mechanics change effects, ideally, may only prefer GSYNC whenever your game is virtually always running below max Hz -- e.g. games like PUBG. Basically games that almost never reach a full 240fps during 240Hz GSYNC.

For the majority of casual and competitive players, this won't matter. But it's something merits being said, given diverse readers... and the top-end eSports players who say they never will game with ULMB or GSYNC -- but there are those who want to read BlurBusters to understand the "right tool for the right job" (Blur Busters has eSports followers too...).

If you't not in tournaments in eSports, nor playing with people who do, then the "lag-mechanics" change effects at the VRR threshold is most likely not applicable to you... For more than 99% of readers this will be completely unimportant, but we don't dismiss the importance of milliseconds...

For such players, who play games whose framerates frequently transition above/below monitor VRR Hz, going 100% VSYNC OFF to prevent the lag-mechanics changes (of going above/below max Hz), or using GSYNC only with games that stay below max-Hz -- can still be useful "right-tool-for-right-job" to such competitive players...