FSO vs. Exclusive Fullscreen

[howiec]

https://www.youtube.com/watch?v=E3wTajGZOsA
It's more specific to the flip modes available to DX12 in Windows though.

That said, beyond subjective per-user reporting, there's no easy technical test to definitively prove FSO is worse than exclusive fullscreen in any/all circumstances and in what respect.

All I can say is, again, as I told @TTT, I subjectively haven't felt a difference in more recent Windows builds, and, further, it has never visibly increased input lag in any of my previous high speed tests.

Ah yes, I remember that old but good video.

Yeah, people need to realize that FSO may or may not be "worse" and depends on various factors like the game itself and other settings (e.g. Gsync on/off, vsync on/off). E.g. true independent flip mode clearly should be just as good as FSE if everything is working correctly.

I think the only way to really tell quantitatively what is better for a specific game and system configuration is to use something like GPU View or PresentMon to compare FSO on/off.

However, in my case with Apex Legends, it's just too obvious that mouse "feel" is much more responsive and I know for a fact that DWM is bypassed in FSE requiring FSO off.

For competitive PVP FPShooters I always aim for lowest latency possible, incl. swapinterval 0, ensure bypass DWM, vsync off and tearing is ok, minimize cmd and other buffers, plus many other optimizations. However, there is a trade-off between lowest latency possible settings and frame rate performance/smoothness (affecting motion clarity) which can clearly affect a person's gameplay performance too.

[jorimt]

However, in my case with Apex Legends, it's just too obvious that mouse "feel" is much more responsive and I know for a fact that DWM is bypassed in FSE requiring FSO off.

It could be somewhat "deduced" how much of an effect DWM would have on mouse input without having to do any complex tests by simply considering it's nature.

DWM is essentially a tear-free compositor, so where games are concerned, that means the worst it can do is increase the buffer. E.g. it doesn't play any part in the rendering of game frames, only their display.

A buffer would increase input lag by a full frame in each instance (as opposed to fractional frames), so we could conclude that if FSO is doing anything to mouse input, it would be in single increments of 1 or more frames, and only after the frames themselves have been rendered.

And as I have said, I have done high speed testing of FSO vs true exclusive fullscreen in the past, and it hasn't shown an increase in average input lag.

There could obviously be numerous other factors earlier in the render chain that may affect mouse input in a more granular manner, but short of full frames of delay, I don't personally see how the DWM could affect it so granularly where FSO is concerned.

That, and I just checked in Overwatch, and FSO allows tearing when the game is set to exclusive fullscreen. This would certainly point further to DWM being fully disengaged when the game window is focused on.

Anyway, just something to consider in place of more exhaustive test data.

[howiec]

A buffer would increase input lag by a full frame in each instance

I'm sure many of us are fully aware of the minimum additional 1 frame latency.

Citing another probably overly-cited article, there's more to it than just the 1-frame latency - you have to consider total (sub/)system-level effects:
https://jackmin.home.blog/2018/12/14/sw ... t-latency/

[diakou]

However, in my case with Apex Legends, it's just too obvious that mouse "feel" is much more responsive and I know for a fact that DWM is bypassed in FSE requiring FSO off.

And as I have said, I have done high speed testing of FSO vs true exclusive fullscreen in the past, and it hasn't shown an increase in average input lag.

There could obviously be numerous other factors earlier in the render chain that may affect mouse input in a more granular manner, but short of full frames of delay, I don't personally see how the DWM could affect it so granularly where FSO is concerned.

Anyway, just something to consider in place of more exhaustive test data.

NVIDIA's "comment"; (It is important to keep in mind that they are most likely referring to Borderless Full screen and not Full screen Optimizations. Fullscreen Optimizations simply turn Fullscreen Exclusive into a "borderless" variant of fullscreen exclusive, but is reported to have "identical performance." as FSE.)

https://www.nvidia.com/en-us/geforce/gu ... tion-guide

We have no way of knowing what more effects the debate and why it feels marginally worse for some people, even though there's a statistically "insignificant" change between FSO and FSE. Resources not being focused entirely on the window maybe?

Either-way, most people do not even know they're running FSO these days, majority of the windows 10 user base got bruteforced by Microsoft as devs were not implementing their own improved flip models for borderless fullscreen, thus they themselves turn exclusive full screen into a borderless full screen without the "overhead/performance decrease" for all the alt-tab goodness and more.

RodroG's capframex analysis of FSO vs FSE vs FSB
https://www.reddit.com/r/allbenchmarks/ ... ullscreen/

interesting all around in all honesty, it's not as mysterious anymore like it was in the past but still confusing enough.

(Also yes, I know I regurgitated information you already know, doing it for readers who don't read entire threads)

[jorimt]

Citing another probably overly-cited article, there's more to it than just the 1-frame latency - you have to consider total (sub/)system-level effects:
https://jackmin.home.blog/2018/12/14/sw ... t-latency/

1 frame "increments."

That article confirms what I stated in my previously reply; the DWM is involved in flipping frames, not rendering them. So the most it can do is increase the buffer by 1 or more frames, and, where games are concerned, drop frames due to it's Fast Sync-like V-SYNC solution.

Thus, if we know that FSO allows tearing (which confirms it bypasses the DWM syncing method that causes dropped frames), and it doesn't show an increase in average input lag when tested against exclusive fullscreen (which confirms it bypasses DWM composition's added buffering), we can also safely conclude the two modes are for all intents and purposes the same from a performance perspective.

NVIDIA's "comment"; (It is important to keep in mind that they are most likely referring to Borderless Full screen and not Full screen Optimizations. Fullscreen Optimizations simply turn Fullscreen Exclusive into a "borderless" variant of fullscreen exclusive, but is reported to have "identical performance." as FSE.)

https://www.nvidia.com/en-us/geforce/gu ... tion-guide

Yup, they're referring to borderless/windowed, which is exactly what FSO would be doing if it weren't behaving like exclusive fullscreen (which it does, as it effectively bypasses additional DWM composition when focused).

We have no way of knowing what more effects the debate and why it feels marginally worse for some people, even though there's a statistically "insignificant" change between FSO and FSE. Resources not being focused entirely on the window maybe?

With the info already available to us, whatever they're feeling is almost certainly not directly due to FSO vs. exclusive fullscreen.

It could be psychological. It could be a variety of other factors separate of FSO that may indirectly affect it in some "mysterious" way. It could be a mix of both, but I'm merely focusing on what IS known and can be ruled out.

I realize it's not popular in these type of discussions to pose an absolute, or to leave no room for some hopeful "possibility," but I've risked offering a potential conclusive answer on FSO vs. exclusive in this respect regardless.

That said, everyone here is obviously free to continue speculation. I just find no need to participate in that further myself, so I'll leave you all to it...

[TTT]

I've been flipping back and forth on Destiny 2 turning FSO on/off and kept gsync off also.

With FSO on, the mouse definitely feels much smoother like some kind of sync is going on. I guess there is something behind this or is it just me?

[ball2hi]

I'd like to chime in here.I know it's pretty cliche to say this in this sort of scenario, but I find I am extremely sensitive to differences.

When I have FSO on, it feels like the game is visually much smoother. It's not an FPS thing, because my framerate stays more or less the same. In some ways, I feel like I have more difficulty aiming with FSO on but in other times I think I just need to re-adjust to the smoother visuals.

When I have FSO off (disable fullscreen optimizations), the game looks very choppy, stuttery even. It's moreso noticeable in very high-motion games that require you to be turning to aim at something that's also moving past you.

I can't really say for certain between the two. I honestly want to say that FSO doesn't cause input lag but rather throws people off because they aren't used to the game looking that visually smooth.

Enabling Xbox Game Bar on the other hand... idk what's up with that. My FPS stays high (+200) but for some reason it feels like the game is visually running at 60hz on my 165hz display.

[TTT]

I'd like to chime in here.I know it's pretty cliche to say this in this sort of scenario, but I find I am extremely sensitive to differences.

When I have FSO on, it feels like the game is visually much smoother. It's not an FPS thing, because my framerate stays more or less the same. In some ways, I feel like I have more difficulty aiming with FSO on but in other times I think I just need to re-adjust to the smoother visuals.

When I have FSO off (disable fullscreen optimizations), the game looks very choppy, stuttery even. It's moreso noticeable in very high-motion games that require you to be turning to aim at something that's also moving past you.

I can't really say for certain between the two. I honestly want to say that FSO doesn't cause input lag but rather throws people off because they aren't used to the game looking that visually smooth.

Enabling Xbox Game Bar on the other hand... idk what's up with that. My FPS stays high (+200) but for some reason it feels like the game is visually running at 60hz on my 165hz display.

This is also how I feel about it, also when you are moving the mouse around quickly it feels snappier and more accurate with FSO off.

[howiec]

I'd like to chime in here.I know it's pretty cliche to say this in this sort of scenario, but I find I am extremely sensitive to differences.
[...]
I can't really say for certain between the two. I honestly want to say that FSO doesn't cause input lag but rather throws people off because they aren't used to the game looking that visually smooth.

Borderless Fullscreen with FSO on Apex Legends leads to minimum additional 1 frame latency. Easily verified using something like PresentMon or GPU View.

With my customized settings and highly-tuned setup, that throws off aim significantly.

[jorimt]

Borderless Fullscreen with FSO on Apex Legends leads to minimum additional 1 frame latency.

On this note, just to be clear on my previous comments to everyone here, when I was talking about FSO vs. exclusive fullscreen, I was talking about FSO on + game set to exclusive fullscreen vs. FSO off + game set to exclusive fullscreen only.

- Game set to "borderless/window" + FSO on = borderless/window behavior (retains DWM composition + Fast Sync-like V-SYNC solution)
- Game set to (exclusive) "fullscreen" + FSO on = exclusive fullscreen behavior (with allowance of adaptive DWM for overlays/alt+tab transitions + tearing with in-game or CP V-SYNC off)

I've seen DSI/DSR scaling quirks sometimes occur with the borderless scenario (vs. borderless + FSO off), however (may have been addressed since though, or was game specific; primarily saw this in Sunset Overrdive which only supports borderless).

Anyway, due to DWM, borderless/window does indeed add 1 frame of latency over exclusive fullscreen:
https://blurbusters.com/gsync/gsync101- ... ttings/10/





For the most responsive experience, borderless/windowed (FSO on or off) must be avoided, and exclusive fullscreen (FSO on or off) must be used instead.

Obviously, exclusive fullscreen isn't always an option though, so borderless/windowed G-SYNC mode is the next best bet, as it appears to bypass DWM in that mode regardless (I will test this again in the latest Windows build for confirmation when/if I receive an LDAT from Nvidia):





Finally, during my original G-SYNC 101 tests, I also tested all combos of Game Mode, Game Bar, and FSO off-record, and found no measurable input lag differences. See:
viewtopic.php?f=5&t=3441&start=10#p26839

For the record, I know my "G-SYNC 101" article only evokes "G-SYNC" when people see that title, but in actuality, it fully covers G-SYNC, double buffer (and to a lesser degree triple buffer) V-SYNC, Fast Sync, V-SYNC OFF (no sync), exclusive fullscreen, borderless/window, and in-game/external FPS limiter scenarios (and across 6 refresh rates).