Does true full screen exclusive still exists in Win11?

[matrox-40]

23H2

1231.jpg

Are you interested in this?

Hi I'm interesed in everything that would allow me get stutter-free iRacing, and would allow me to undestand, why sometimes its smooth ans sometimes not, while I don't change anything.

[1000WATT]

Check PM.

[Gias]

Hello.

In Win11 24H2, windowed vs full screen, despite all scenarios were showing "hw composed: independent flip" Full screen was much more smooth than windowed. Windowed was just unplayable, for sune no placebo. Even recorded with 240FPS camera and was analyzing in slow-motion.

hmm that doesn't make any sense to me if everything else was equal like presentation interval, vrr...

your fullscreen mode there is technically windowed also (in a higher z-band than the desktop, but that normally doesn't translate to better performance compared to a windowed/borderless game on the desktop z-band unless something gets on top of the game and puts it in a composed flip presentation, which doesn't appear to be your case here since you keep getting "hardware composed: independent flip" for presentation...)

I swiched to Win10 22H2 to see how it goest there, as its generally more smooth for me (I did tons of experiments before I started thread here), and here are findings:

I discovered something interesting that in windowed mode presentation model changes by itself: every several seconds I can see model name changes in that RTSS overlay, but change is so fast you cannot read whats there. I recorded video and then analyzed frame by frame and it was like below. In this situation there is stuttering.

huh yikes... if it's actually constantly changing between "hardware: independent flip" and "hardware composed: independent flip" then it could be because the game is somehow constantly changing between plane 0 and plane 1+, but a constant change like that isn't normal...

Also I found, that sometimes in windowed mode game will start as "composed: flip" and this is stutterfest, also what is very misleading in this scenario windowed gsync seems to be activated, monitor Hz fluctuates, g-sync indicator is on but game play is stuttermess.

20241020_195041.jpg 20241020_195022.jpg

keep in mind that rtss has a 3 seconds reporting lag by default for the data it displays from intel's presentmon (this includes the presentation model reading); basically because intel's presentmon has a delay for that and to allow for smooth scrolling of data...

gsync shouldn't be working when a game's got composed flip for presentation (unless you have the "enabled for windowed and full screen mode" option for gsync enabled in the nvidia control panel... but, then again, that option in practice may not work great since it may not sync vrr properly to the game and vrr may sync to other stuff you may not want it syncing to...)

In full screen I see constant "hw composed: independent flip", however when I turn on menu start, model changes

20241020_225459.jpg

this would indicate then that your game is on MPO plane 1 but you don't have 2+ MPOs available to retain independent flip presentation when something gets on top of your game... hence the game falls into the composed flip presentation when you bring up the windows start menu over the game.

so its different behavior vs Win11 24H2. However, when "hw composed: independent flip" is there permanent in full screen, I can get it very smooth with g-sync & frame cap via RTSS or ingame cap.

You said "independent flip presentation" = as good or better than FSE. Is this :
"hw composed: independent flip"
or
hw: independent flip" or any of these ?

both.

"hardware: independent flip" and "hardware composed: independent flip"

or either one of those.

that's actually why i normally say "independent flip presentation"

those should perform the same. those bypass the windows desktop manager and allow the game to get more out of the cpu/gpu compared to a composed flip presentation etc, plus an independent flip presentation may use waitable swapchains to reduce latency further in certain scenarios (and waitable swapchains aren't supported in FSE).

What is the difference ?

"hardware: independent flip" indicates the game did not take ownership of the screen (either because it's just borderless windowed or because FSO/eFSE was applied), but it is still swapping the displayed surface every frame (via buffer pointer swap/flipping), independently, with the same efficiency as fullscreen exclusive (hardware: legacy flip).

"Hardware Composed: Independent Flip" is the same as the above, but also indicates the app has been granted at least 1 hardware overlay plane (MPO plane).

MPO (multiplane overlay) planes are basically additional hardware scanout planes enabling the gpu to take over composition from the desktop window manager, thus allowing the game to retain independent flip presentation in more scenarios.

and it's normally needed to have at least 2+ MPO planes assigned and available to the display the game is running on if you want MPOs to save you (keep your game using an independent flip presentation in more scenarios).

also typically it's "hardware: independent flip" because the game is on plane 0 (but the system could still leverage the MPOs and the game may retain independent flip presentation in certain scenarios if 2+ MPO planes are available...)

the presentmon service / session may also get stuck, but on my end that has been very rare... unless i had multiple different apps using stuff from presentmon at the same time (like sk, rtss, nvidia's overlay, intel's console presentmon etc). then you may need to stop and restart the presentmon event trace session(s) in performance monitor to get the presentmon data working again...

and it's "hardware composed: independent flip" if the game got on plane 1 or higher (and if it's on plane 1 and there's only 1 MPO plane available, then the game would still fall into the "composed: flip" presentation that involves dwm composition if some external window is on top of the game or if there's a resolution mismatch between the game's window and the desktop etc)

plane 0 is also the dwm's plane (the dwm's swapchain); however, even if the game is on plane 0, dwm composition can still be bypassed if the game has engaged directflip/independent flip optimizations (with or without MPOs if certain criteria has been met...)

the following documentation from microsoft goes over some of that if anyone here cares:

https://learn.microsoft.com/en-us/windo ... quirements

https://learn.microsoft.com/en-us/windo ... flip-model

I'm still manipulating a lot, so this is just in-between update. My target is to make is working smooth in tripple screen configuration.

Besides presentation model, do you think can there be some explanations why in Win11 I cannot get it smooth in any scenario, while in Win10 yes ? I was trying to replicate all settings from Win10 to Win11 and Win11 stutters.

i'm not sure what all is different between your different tests. it could be a number of things...

i'd still check the gpu clocks and enable nvidia's prefer maximum performance option in a driver profile for the game in the nvidia control panel. you may also want to test with only one monitor connected if you haven't already and try closing other apps you may have running in the background...

honestly i suspect your windows install is broken in some way or something else you got running is causing a conflict somehow.

sorry i don't have this game to test on my end. i'd also test some other games if i was in your situation though.

[RealNC]

Since this is a DX11 game, one thing you can try is to force the flip model with RTSS. Create an RTSS profile for this game, then quit RTSS (also quit Afterburner if you use that too so it doesn't auto-relaunch RTSS) open the profile in a text editor. The profiles are located in the "Profiles" folder in the RTSS installation folder.

Add this to the profile:

Code: Select all

[OverrideDXGI]
ForceFlipModel = 1
ForceAllowTearing = 1

Make sure the game is set to borderless fullscreen, not exclusive fullscreen.

At the end of the day though, when you have multiple monitors, expect problems. MPOs usually don't work correctly when having multiple monitors.

[matrox-40]

Thanks for all inputs.

In the end, I cannot connect microstutters to presentation model. Initially I thought I have something, but after several days all my finding were miss.

Its only iRacing to give such trouble, I have other sims like AC1, ACC, AMS2 and all they are supersmooth, even with just VSYNC=ON, not yet invoking G-SYNC. Of course with G-sync they are all working OK, so G-sync is somehow "needed" more in situations, where I get FPS dips. As my PC can hold constant 100 or 120 FPS, VSYNC is good enough for me.

With all being written here, this is really complicated subject. I understand new technologies\approach coming in, but it seems just FSE is taken out of Windows. I mean that brutal FSE that will not allow volume bar or start menu going on top of game. Its not there in the name of what ? Fast alt+tab? Lower latency ? Adding overlays? Yes its all good, but not fundamental.

I'm not sure it would solve my problems, but at least I wanted to give it a try with this old title. Its known for kind of ancient approach (and possibly code).

Greetings.

[matrox-40]

Since this is a DX11 game, one thing you can try is to force the flip model with RTSS. Create an RTSS profile for this game, then quit RTSS (also quit Afterburner if you use that too so it doesn't auto-relaunch RTSS) open the profile in a text editor. The profiles are located in the "Profiles" folder in the RTSS installation folder.

Add this to the profile:

Code: Select all

[OverrideDXGI]
ForceFlipModel = 1
ForceAllowTearing = 1

Make sure the game is set to borderless fullscreen, not exclusive fullscreen.

At the end of the day though, when you have multiple monitors, expect problems. MPOs usually don't work correctly when having multiple monitors.

Hello let me update on subject, as I have new findings.

You were right: more monitors = more trouble and this might be a case. Nevertheless when I started subject, I was just experimenting with single monitor. Once I got it sorted (understood, I mean flip model story), I went to tripple configuration analysis.

So below relates to tripple monitor config, windowed borderless. Center monitor is main monitor, I wanted to get g-sync working at least (in fact: only) for center monitor, side monitors are not important to me in terms of smoothness - they are mainly to ensure visibility what happens around my car.

What I have found, is that game randomly uses one of two modes : "composed flip" (that sutters a lot) or "hw composed flip" (which is best on my side, maybe not supersmooth, but playable). I mean game will always open as "composed flip" and in 2-3 seconds it switches to "hw composed independent flip" when its foregound, in background its "composed flip" but for every change back\foreground it changes as expected. Surprisingly, other times it will open as "composed flip" and will stay like that, despite being foreground.

I have noted that in situations when "hw composed independent flip" was engaged, I close game, shut down PC, and with anything changed next day it will not be "hw composed independent flip" anymore. It will stick to "composed flip", which stutters. That is answer to my question, how on earth it can be one day game is stuttering, other not, I was even cloning whole disk to have it protected, but then shockingly clone also was stuttering...

If game sticks to "composed flip" if I relaunch it or reboot PC > it will stay like that. To get it back to "hw composed independent flip" I need to : disable g-sync in NVCP > apply > enable g-sync in NVCP > apply > and it helps.

Advised method to enter code in RTSS:
[OverrideDXGI]
ForceFlipModel = 1
ForceAllowTearing = 1
doesnt work. Game doesn't react to this forcing. Is there anything else I could try ?

Here is screen how its shown in RTSS by using presetmon plugin.

hw flip to composed.png (41.21 KiB) Viewed 4556 times

Another day, when I had "hw composed independent flip" and I shut down PC, I was ready next day to start with procedure to disable>enable g-sync, as it seems to be mysthical cure. I mentally accepted this will be just starting procedure of iRacing, as it helps on my side. But on some days its not needed. It will keep "hw composed independent flip" after reboot or shutdown.

What I described above reminds me one thread found, where developer was searching why his app behaves different in terms of presenation model at three different computers with same Windows & GPU driver version.
https://stackoverflow.com/questions/725 ... -swapchain

I still don't understand if this is game bug, or windows bug or what. But at least now with presentation model displayed, I can see whats going on and eventually react.

In the end I want to tell its still not suprsmooth as I would like to have it (and how it should be compared to other titles, while at 120FPS or more). I have doubts if internal frame pacing is correct, but at least game now is playable, not annoying.

Regards,

[NJC-Developments]

@matrox-40
Hi, new user here and I've found your post after weeks of head scratching debugging iRacing sutters on Windows11 24H2. I have tried everything i can think of (except disable MBO, which lead me here) but wanted to say my symptoms are exactly the same as yours. Assetto Cora runs buttery smooth, iRacing runs fine one time, then stutters occassionally the next. No changes made between sessions, and no graphic changes either. In fact, graphical settings (lowering demands on CPU/GPU) doesn't make any difference to me.

I use a single LG OLED (CX) screen using G-SYNC and FPS capped in game to 117fps. Other games G-SYNC doesn't cause any problems. FPS can dip to 80 and i would hardly notice it. In iRacing, when stutter happens, I can see in-game performace meter momentarily drop to 110fps, but stutter feels like its running at 40fps! I'm now wondering if this change in presentation mode is something i too am expericing.

I dont have RTTS (or even know what that is yet) but happy to learn and try and capture my experiences if it helps?

[matrox-40]

Hi after all possible things I checked and done to get it smooth, my final conclusion is that there is some glitch in iR code itself. I think there is hope when they release DX12 version.

[NJC-Developments]

I was pretty much reaching the same conclusion myself. Hopefully this year it happens, and we can all just enjoy racing again. I will update this if I do find something that fixes it in the mean time.

[matrox-40]

Yes you are correct I believe and I wish something happens from iRacing side to get it sorted. After several months since I started topic here, iRacing also keeps updating, Windows updates, nVidia updates...there is almost no possibility to make long term tests with the same stable version. Currently I see still microstuttering but it's not exact the same as it was. I feel that distant objects (like trees hills etc) are stuttering as earlier, but for example closer objects like signs fences cones are more fluid. But still it may vary day-to-day.

But just answering topic question, I think there is no sense anymore to expect/desire full screen exclusive anymore with the way Windows progresses. When I learned all that stuff about presentation models, variety of options that are there, it's a pity it's not just all growing next to stabilized FSE. For your well being they just decided true FSE is not there (windowed borderless imitates FSE), they develop much more complicated scenarios which many users possibly don't need or even don't want.