Stutter/judder/ghosting without frametime increase or framerate drop (likely not just in VR)

[firdimigdi]

The issue

Singleplayer missions (career, AQMB, scripted campaigns - generally missions with adequate AI activity [nothing extreme]) are juddery despite frametimes being low and the framerate being locked to the maximum for the HMD (I have tried both 60 / 90fps on a G2 and 80fps on a Rift S). This is something that has started in some update near January 2022. It is only visible in the HMD, the desktop mirror does not exhibit this.

 

Recording a track while this is occurring and playing it back results in a completely smooth playback. It is also detectable if you decelerate time in mission but somewhat less so if you accelerate time in mission.

 

What does it look like

It's as if the rendered frames are reaching the HMD slightly out of sequence - this happens for both eyes, so it's not an inconsistency between left and right.

 

Easiest missions to see it in

Fortress over the Volga, 1st mission

Burning Steppe, 3rd mission

Or the missions I've attached in this post

 

Roll left right and keep your eye on the horizon out of the side of the canopy and it will become very obvious; alternatively get close to a plane and barrel roll around it while keeping it in view.

 

The hardware this was tested on

5900X/32gb 3600Mhz/RTX 3080Ti

i9 9900K/32gb 3600Mhz/RTX 2080

Reverb G2

Oculus Rift S

 

The software tried

 

All machines are running Windows 10, latest nVidia drivers - tried also rolling back to the 496.xx generation for testing's sake.

 

• Native unadulterated SteamVR.

SteamVR chart from the G2 while the game is juddering

Spoiler

 

And from the Rift S

Spoiler

As is clear from both charts there are no dropped frames, no synthetic frames and it operates well within the threshold for a maxxed out and stable framerate.

 

• OpenComposite (OpenXR branch for the G2 and the original for Rift S) - this yields even better/smoother frametimes, yet the issue is still very much present.

 

I tried running the lowest of the low settings. I tried a completely clean install of IL-2 despite there not being any mods installed when testing. The judder always occurs as long as there are enough AI planes/vehicles to induce it.

 

Also, I swapped the HMDs between both PCs and got the same result.

 

What it is not

- Motion smoothing/reprojection, I kept it disabled.

- A discrepancy between left and right eye rendering, it is still perceived with either eye closed.

- Game settings dependent, it happens even with the lowest graphics settings and at 50% the HMD's render resolution.

- In-game weather related, it can happen with clear cloudless skies.

- Map-related, can happen on any map.

- Distance related, as long as the AIs are spawned they will have an effect.

- Framerate drop, the frames remain at max.

- CPU overload, CPU usage remains somewhere under 50% at all times.

- Thermal throttling, both PCs have liquid cooled CPU and GPU and never reach their thermal limit.

- Dependent on CPU architecture, occurs both on Intel and AMD.

- Dependent on HMD ecosystem, happens just the same with Oculus as with a WMR headset.

 

After several rounds of testing:

It's the presence of AI-controlled airplanes that is causing this the most, even if they are not active/flying and just parked.

Below is a set of two MP dogfight missions: in one there are 33 ground vehicles moving along a road and it does not exhibit ghosting/stuttering, in the other there are 18 planes just parked on a runway and the issue is triggered.

 

stuttertest - two missions vehicles vs planes.zip

 

And yet another MP dogfight mission making it even easier to test:

 

stuttertest - plane spawner on flare.zip

 

Shoot a flare and a FW190 will spawn and start circling - doing it this way I saw that it actually starts much earlier than I thought it would, I started noticing micro-stutter at about 10 spawned planes when playing this mission locally.

 

If you play the above mission in multiplayer, hosted on another PC, they will be smooth.

 

Other posts mentioning this issue

 

And some more commentary here:

Original post:

Spoiler

I've been trying to figure out an issue I keep seeing in SP scenarios, or Dogfight missions in a server hosted in-game with multiple AIs and especially with ground units. This is most easily perceived in VR (note: there is no motion reprojection/smoothing involved).

 

Steps to reproduce easily:

1. use the Easy Mission Generator to make a Dogfight ground attack vs vehicles mission (it doesn't have to necessarily be such a mission as the generator will add ground units usually on the map anyhow, but it's easier this way) - alternatively AQMB can be used but there's extra steps to get the _gen mission modified to use in step 4

2. start a server in the game and load this mission

3. observe that when there are multiple ground unit AIs active that while the frametimes remain low and framerate remains high there is a very obvious reduction in the fluidity of motion - no dropped frames or even synthetic frames are shown in SteamVR/fpsVR's performance overlay

3b. easiest way to notice is if you are in a plane with high roll rate and flying over a road - keep your eyes on the road and do quick rolls left-right, observe how the road can appear slightly ghosted

4. host the exact same mission on another computer (dserver) and everything runs perfectly smooth

 

Note: you do not need to be near the AI for it to occur as long as they're spawned and active somewhere on the map

 

I can also provide much simpler missions than those generated by the Easy Mission Generator which reproduce the exact same result as above and it is also something that occurs with AQMB missions, career missions and scripted campaign missions.

 

One can rush to say "yes, ok, it is taxing your computer less since you are running it on another PC" but this should surely manifest itself in the framerate actually going down and there being excessive CPU/GPU usage, which is not the case, the framerate indicator in game remains locked at 90fps (as it does in any external monitoring utility) and the frametimes remain way below 11ms which is the threshold for maintaining 90fps.

 

The hardware involved is a 5900X/32gb 3600Mhz/RTX 3080Ti - the secondary machine used with dserver to test step (4) uses a "lowly" AMD 3700U.

 

I guess it's worth mentioning that weather conditions or reducing graphics settings have no bearing on this - it is always there as long as AI ground units are involved.

 

Edited August 16, 2022 by Firdimigdi
typos, formatting, retitled

[AEthelraedUnraed]

I've got the same issue when on SteamVR. Using OpenComposite, everything is smooth.

[firdimigdi]

On 5/22/2022 at 11:28 PM, AEthelraedUnraed said:

I've got the same issue when on SteamVR. Using OpenComposite, everything is smooth.

 

Unfortunately that doesn't seem to be the case on my end. In fact it was the first thing I tried when OpenComposite's OpenXR branch reached a stable enough state. There is a difference, maybe a slightly less apparent loss of fluidity but other than that it's still a thing. And it's totally baffling because by all metrics it should not be happening so somehow it points at something occurring in the game's AI-controlling thread.

 

In fact, and I should add this to my initial report, you don't even need to be in close proximity with the AI for the issue to occur as long as it's active; it can be half-way across the Kuban map - I'll try to knock up the simplest mission which manifests this and put it here as a test case.

 

Edited May 25, 2022 by Firdimigdi

[firdimigdi]

Here's a MP dogfight mission thrown together from groups I had lying around with the minimum number of AI I found was needed to induce the stuttering; it's a mix of ground/air/naval units (but still way less that I found can be active in AQMB/campaign/career misions). You get to spawn with a thin strip of land on your right which makes it easy to spot the stuttering when you start rolling left-right - it's subtle in this case but adding some more AI (as you'd find in SP missions/scripted campaigns/career) makes it way more pronounced.

 

If testing with VR (where it becomes very apparent) it's best if you do not have any motion smoothing/reprojection activated to get the full effect, although with motion smoothing on it still has a perceptible stutter.

 

It's not optimized or anything (some units might activate/deactivate based on enemy proximity but the rest should remain) but if I host it on a separate server it doesn't exhibit any of the stuttering I'd get when hosting it in-game. If I reduce the number of AIs (or just limit it to airborn AI) then it becomes smooth once more.

 

Another extremely easy way to test using official content is the third mission of the Blazing Steppe campaign - turn on autopilot, set accelerated time and as soon as the IL-2's take off and are about 1-2km away from their airport return to normal speed and take control of the plane then try rolling left right - even when at a distance from the combat sector there's a perceptible stutter/ghosting/something and the motion is not fluid. This persists during the whole mission and all the while the reported framerate does not drop. I even tried it with much reduced graphics settings (low preset, everything off) and it's still the case (which isn't surprising) and CPU occupancy is way low.

 

I honestly think this is an issue that has snuck in there relatively recently as I never remember something like this from say a year ago or so when I found myself playing singleplayer modes (as is obvious: this isn't apparent in multiplayer). Some cost in performance is expected with multiple units active but normally that also translates to framerate dropping or something similar (like time-dilation in extreme situations), this does not.

 

stuttertest.zip

Edited May 27, 2022 by Firdimigdi

[dgiatr]

Υes exactly the same thing happens to me, i see little difference if i use steam vr or openxr. i get 75-90 fps but this kind of stutter occurs lately, i dont think it happend in the past.

[firdimigdi]

To further illustrate the point and put some metrics in to it - this here is a SteamVR frametiming diagram using a G2 while running the Blazing Steppe mission 3 mentioned above:

 

As you can see, apart from minor spikes the rest is perfectly acceptable and keeps running locked at 90fps - despite that, what you see in the headset is not smooth at all. It's as if the world outside of the cockpit is rendering at 60Hz instead of 90.

Edited May 31, 2022 by Firdimigdi

[firdimigdi]

And just as a test, here is the same Blazing Steppe mission 3 for reference using a Rift S, with much lower resolution as expected and naturally much much lower frametimes - yet again the motion inside the headset is not smooth at all:

 

 

Unfortunately I have no idea how to show this as there's no way to actually record a useful video of it.

 

I suspect that this issue started somewhere in the 4.70x updates in January or so, it is not weather/cloud related despite those being the major points of said updates but it is clearly related to AI controlled units, however I cannot be sure 100% since I wasn't playing singleplayer much at the time.

 

EDIT:

Another point of interest is that a recorded track of the mission above is smooth when played back, but of course that is no surprise as the AI units aren't active in the track.

 

Edited May 31, 2022 by Firdimigdi

[horendus]

I can confirm this issue as well, feels like bad frame timing but frame tools show consistent pacing, only when playing scripted campaigns.

 

Im no game engine developer but it seems like the engine is sending this micro stuttering mess to the GPU to render which is why the frame timing tools dont show any slow downs. The system is infact reaching those required render times, complete with the JANKED pacing its being told to include. 
 

“Here you go mr. GPU. I want you to render these next 90 frames for me….but make it feel as if its 45 frames please. Perhaps pin the frame update timing to the completion of this overly intensive AI loop”

[firdimigdi]

Original post updated for better formatting and clarity in a format more consistent with a bug report.

[Regingrave-]

22.05.2022 в 21:36, Firdimigdi сказал:

After several rounds of testing:

It's the presence of AI-controlled airplanes that is causing this the most, even if they are not active/flying and just parked.

Below is a set of two MP dogfight missions: in one there are 33 ground vehicles moving along a road and it does not exhibit ghosting/stuttering, in the other there are 18 planes just parked on a runway and the issue is triggered.

Even when parked or idling, planes and vehicles are nonetheless fully simulated, that's why there are special static planes and vehicles (which in game terms basically are buildings) to add into missions instead of complex objects. Simulation, exceptionally complex for our planes, requires much of the computing resources — there is not much to be done here as long as our game is a simulator, not an arcade. And optimization of industrial grade aircraft simulation for smooth performance in multiplayer or VR is one of foundations of development of the game itself, not a simple feature one could requre to improve between other tasks.

[firdimigdi]

2 hours ago, Regingrave said:

Even when parked or idling, planes and vehicles are nonetheless fully simulated, that's why there are special static planes and vehicles (which in game terms basically are buildings) to add into missions instead of complex objects. Simulation, exceptionally complex for our planes, requires much of the computing resources — there is not much to be done here as long as our game is a simulator, not an arcade. And optimization of industrial grade aircraft simulation for smooth performance in multiplayer or VR is one of foundations of development of the game itself, not a simple feature one could requre to improve between other tasks.

 

Agreed - however, in campaigns and career missions there are often about as many planes and vehicles active, if not more, and there's not much we can do about that as players. And this was not an issue we had before either sometime in December 2021 or January 2022. As described in the report I posted, you can see this starting to happen with less AI planes than 18. When there is a CPU overload it generally manifests itself in lost frames, or at least synthetic frames, or time dilation when it is extreme. However none of that applies in this case. I'm not talking about a simple hiccup, or a drop of framerate or anything of the sort; this is continuous.

 

The only reason I added them as parked was as part of testing to see if it was maneuvering or dogfighting logic code or something of that sort that triggered it.

 

All I ask is that someone tests it out in VR and if that is how it's supposed to be and there's nothing to be done, so be it. But I believe that what we are seeing is not the usual "CPU overload" or anything of the sort. MP works as smooth as can be expected.

Edited June 17, 2022 by Firdimigdi

[gydaveb]

2 hours ago, Regingrave said:

Even when parked or idling, planes and vehicles are nonetheless fully simulated, that's why there are special static planes and vehicles (which in game terms basically are buildings) to add into missions instead of complex objects. Simulation, exceptionally complex for our planes, requires much of the computing resources — there is not much to be done here as long as our game is a simulator, not an arcade. And optimization of industrial grade aircraft simulation for smooth performance in multiplayer or VR is one of foundations of development of the game itself, not a simple feature one could requre to improve between other tasks.

I think the thing for me is that this is a new issue. A few months back when playing single player career missions, I wasn't getting any of the stutters or ghosting effects currently present in VR. Something changed, and with that change came some pretty bad performance / visuals. What changed? Beats the heck out of me, I'm not a game developer... but logic would say the aircraft / vehicle count present in career missions has probably been consistent, so is this really the problem?

[Regingrave-]

3 часа назад, Firdimigdi сказал:

And this was not an issue we had before either sometime in December 2021 or January 2022

That's a time window too big to make any clues out of it, we've added too many things during this time to say it's some particular one that causes problems.

I'll consult with the team on that again (but not earlier than the second half of July due to vacations), there can be something, but most probably this is a result of general development of the project and related growth of computing demands.

 

[firdimigdi]

1 hour ago, Regingrave said:

That's a time window too big to make any clues out of it, we've added too many things during this time to say it's some particular one that causes problems.

 

Unfortunately I was playing multiplayer around that time, once every two weeks maybe, so I didn't notice, otherwise I'd have of reported.

@Angry_Russian however puts it somewhere between the release of the cloud update and a subsequent hotfix for it if that helps narrow it down to something specific:

 

1 hour ago, Regingrave said:

I'll consult with the team on that again (but not earlier than the second half of July due to vacations), there can be something, but most probably this is a result of general development of the project and related growth of computing demands.

 

Great, thanks - that's all I ask for really that this be looked in to because IMO it doesn't bare the hallmarks of a "general development of the project and related growth of computing demands". Since there is no discernible increase in CPU usage, no dropped frames, no frametime increase, no time dilation, none of the usual metrics are out of line and it's not showing up on what you would call weak PCs either and to top it off as illustrated above it even occurs when the HMD is set to a very low resolution output as well. If anything else in fact in all other regards, on the computers I've tried it, IL-2's performance has actually been increasing (relative to the visual result) with each of the recent patches.

 

Edited June 17, 2022 by Firdimigdi

[Regingrave-]

1 час назад, Firdimigdi сказал:

@Angry_Russian however puts it somewhere between the release of the cloud update and a subsequent hotfix for it if that helps narrow it down to something specific:

Hmm, he mentions the hotfix for the new clouds. This could be some clue, we'll see if we could dig something out of it. Thank you!

[firdimigdi]

@Regingrave I tinkered a bit more on this issue and decided to monitor the game threads to see what is taking up the most of the time.

 

In all of the below cycles delta is calculated per second.

 

This is without any AI present:

 

The HUD is hidden and there are no elements of the UI visible, yet the busiest thread is the Scaleform rendering thread (19368 in this case). This is the UI rendering thread, for comparison, in the menus the cycle delta for it is around ~800,000,000. Note thread 8056 - this plays a big role in the next section. Obviously, this is also what it looks like when joining multiplayer.

 

This is with AI present:

 

Two threads both using more than all the rest of the game threads with cycle delta balooning up to that. Both threads (19368 and 8056) are Scaleform rendering related despite the UI not being visible. I think this is a good clue as to what is going on - the presence of AI can cause the GUI rendering thread to increase its calls by a substantial degree.

 

I honestly think, unless I am mistaken in the assumption of what is creating this increase in calls, this is probably the cause of what we are seeing and why the problem is not alleviated by reducing graphical settings or resolution.

 

Edited July 6, 2022 by Firdimigdi

[Regingrave-]

4 минуты назад, Firdimigdi сказал:

I think this is a good clue as to what is going on - the presence of AI can cause the GUI rendering thread to increase its calls by a substantial degree.

That can be very useful, thank you for your research!  We'll take it into consideration, but Scaleform is a quite heavy framework as it is, this is a longstanding limiting factor for us, as we can't really do anything about it.

[firdimigdi]

Just now, Regingrave said:

That can be very useful, thank you for your research!  We'll take it into consideration, but Scaleform is a quite heavy framework as it is, this is a longstanding limiting factor for us, as we can't really do anything about it.

 

Yeah, I've worked with Scaleform in the past and it needed a lot of tinkering to get it to behave (especially on game consoles). I saw there was a job opening for a Noesis dev on the russian forum, so hopefully someday in the future IL-2 can be disentangled from that monster of middleware.

[firdimigdi]

@Regingrave, sorry to tag you twice about this but I think it might be of interest and probably points to something more insidious than simply Scaleform being the heavy middleware that it is.

 

Using very sleepy (https://github.com/VerySleepy/verysleepy) I obtained some more information.

 

Over a sampling period of 38 seconds, this is the amount of CPU time spent exclusively in various simulation calls (I highlighted a few initially until I realized that all the simulation calls come from the RSE module):

 

 

And this is the amount of CPU time spent exclusively on Scaleform (with no UI elements on screen):

 

 

The other calls make perfect sense since there are more assets being drawn, so you'd expect to see an increase in CPU time from things like the gpu driver (nvwgf2umx) or d3d. But you wouldn't expect Scaleform to cost more than all that.

 

And here from a 38 second sample of a player controlled I16 flying alone:

 

Again compare how much time the gpu driver and d3d is using  - considering this was captured at higher altitude so the overall scene was not as graphically busy as well it's on par with the previous capture that had AI planes and more assets present.

 

Not entirely sure, but it seems to me like as if the simulation engine is being queried by the UI for each plane as if they were the player-controlled plane. If that makes sense. Which could be an explanation of the calls increasing to such an extent.

 

Anyhow, I hope this isn't a red herring from me misinterpreting the data (although I did corroborate it with other profiling tools, it is always a risk when profiling without available debugging symbols) and it proves helpful.

 

Edited July 6, 2022 by Firdimigdi

[Tchandour]

For what it's worth, I've experienced this very issue already in August 2021. Just mentioning since I saw you guys write about when it may have started to occur to people.

[HotCarl73]

Just add my observations to this thread: Highish Spec PC (R9 5900/RTX 3080/32 Gigs DDR4 3200/980 NVME), overclocked or not overclocked, running medium/low graphics no AA, 90% render scale/ 85% NIS scaler. And of course at 100% on both also. All AI planes and terrain ghosting at all times in singleplayer despite max frames and low frametimes. Does it on open composite and steamvr, Reverb G2V2 and valve index. No ghosting in multiplayer. 

[Angry_Russian]

Did anyone try 4.707, is the issue still there? I've uninstalled the game because of this a while ago, but hoping to return when it gets fixed..

[firdimigdi]

13 hours ago, Angry_Russian said:

Did anyone try 4.707, is the issue still there? I've uninstalled the game because of this a while ago, but hoping to return when it gets fixed..

 

Still there.

 

Sadly, I don't think there's much hope of it being addressed, especially now with the BoN map coming up. And if my assumption that it is related to the UI middleware holds true then chances are that it has been chalked as "nothing we can do/not worth the development time - wait until we've changed the UI to use Noesis".

[firdimigdi]

@Regingrave , I know everyone's busy with BoN and this is probably low priority, but some more notes on something I noticed about this issue:

 

In a test scenario where it happens from the get-go (the aforementioned scripted campaigns for example) if you switch to the external chasecam and do a few rolls the scenery around your plane renders smoothly. This camera does of course have smoothing, but even when banking through multiple quick rolls the game world is perfectly smooth - however: your own plane's motion is not, you can see it stuttering ever-so-slightly much like the AI planes do. If you switch to the cockpit camera then everything, gameworld and other planes are stuttery when banking and it seems to follow the stutters of your own plane as seen from the external camera, so basically the symptom we see is because the cockpit camera is attached to a plane that is stuttering, not that the game's rendering of everything is stuttering and this is alleviated somewhat when viewing from a camera that is not rigidly attached to the plane.

 

The above obviously is simply a symptom of whatever is causing the locking of the simulation thread; and after my previous foray in to the activities of the process I am pretty much convinced it's something to do with the excessive Scaleform calls.

 

Speaking about Scaleform - I also noticed that the game has Scaleform's AMP port open (7534). AMP is the profiling/debugging tool for Scaleform; by default it outputs a stream to the above UDP port. Sadly the version I have from a previous project I worked on is newer then the one IL-2 uses and thus I was unable to get a usable feed from it. However, if the game's UI is indeed outputting profiling/debugging data to that port perhaps it would be a good idea to disable it.

 

Edited August 30, 2022 by Firdimigdi

[firdimigdi]

Unfortunately my fears that the scenarios offered by BoN would exacerbate this issue have not been unfounded. Sadly AQMB generated missions in BoN even with "weak" opponent presence are enough to trigger this - perfectly solid 90fps on a G2 or 80fps in a Rift S and yet the motion is not at all smooth.

 

Already there is at least one report from a player using screen with v-sync enabled that I know of which describes a similar result and shows this isn't just a VR issue, although it likely takes a bit more effort to induce it:

 

Is this something that has been acknowledged or is playing anything more complicated than 8vs8 dogfight in singleplayer a write-off for now?

 

Edited September 21, 2022 by Firdimigdi

[chiliwili69]

On 5/22/2022 at 8:36 PM, firdimigdi said:

Below is a set of two MP dogfight missions:

 

I see those missions are in MP. I have tried to put them in the Mission folder but they don´t appear in IL-2.

I tried to delete teh msbin file as well but they don´t appear.

I hvae tried the Fortress Volga 1st mission and I see the sttuters you mention. They are present with my Index at 80 Hz.

[firdimigdi]

10 hours ago, chiliwili69 said:

 

I see those missions are in MP. I have tried to put them in the Mission folder but they don´t appear in IL-2.

I tried to delete teh msbin file as well but they don´t appear.

I hvae tried the Fortress Volga 1st mission and I see the sttuters you mention. They are present with my Index at 80 Hz.

 

Yes, they are multiplayer missions so it is easier to respawn and try out stuff and more importantly so one can compare the performance of hosting the mission on a different PC (where, as noted above it plays perfectly smooth). To get them to work you need to put them in the multiplayer/dogfight folder and start a multiplayer game (you can deselect "visible to the internet"). They indeed might need resaving since they were made with an older version of the editor, if that is a problem I'll do that later on today or tomorrow and upload them here.

[chiliwili69]

1 hour ago, firdimigdi said:

To get them to work you need to put them in the multiplayer/dogfight folder and start a multiplayer game (you can deselect "visible to the internet"). They indeed might need resaving since they were made with an older version of the editor, if that is a problem I'll do that later on today or tomorrow and upload them here.

 

OK thanks, I have never entered in the Multiplayer arena. I have been always SP.

 

With the SYN_Vander benchmark, that was created about 2 years ago, I just post the mission zip file without the .msbin file, so this was solving the problem for newer versions. I don´t know if MP mission works in the same way.

 

If you are going to recreate the mission you can use the I-16 plane, so we don´t need to open the cabin to see the stutters better. Also, the I-16 is the best plane to do quick roll.

 

Yesterday with the Fortress Volga 1s mission I was also testing the Index at 120Hz mode with the lower settings and just 80%SS, so I had very low frametimes and solid 120fps. But the sttuters remains there.

I noticed that after flying around for 10 minutes in the Fortress volga populated airfield the planes disappeared from the airfield. But the sttuters remains there.

And the sttuter only apear in the landscape or planes around when rolling. Only when rolling. If the plane is steady and you turn your head left-right or lean left-right the landscape is perfect smooth.

And as said above, if you put the exterior view of the plane, everything is smooth. It is only when sitting inside the plane and rolling left-right.

I think this has to do with some kind of synchronization issue between the plane roll angle and the headset orientation angles. It is weird that this is affected by the number of AI objects in scene.

[firdimigdi]

24 minutes ago, chiliwili69 said:

I think this has to do with some kind of synchronization issue between the plane roll angle and the headset orientation angles. It is weird that this is affected by the number of AI objects in scene.

 

Precisely why I have been pining on about this, as this proves that the motion could be made smooth enough so the landscape is not stuttering like crazy even if the planes are a bit jittery and the experience would be instantly much more pleasant.

 

EDIT:

Also it is why I think it's the physics/simulation thread that is lagging which causes all planes (including your own with the cockpit camera attached to it) to have jittery motion.

 

24 minutes ago, chiliwili69 said:

With the SYN_Vander benchmark, that was created about 2 years ago, I just post the mission zip file without the .msbin file, so this was solving the problem for newer versions. I don´t know if MP mission works in the same way.

 

If you are going to recreate the mission you can use the I-16 plane, so we don´t need to open the cabin to see the stutters better. Also, the I-16 is the best plane to do quick roll.

 

Sure - it would make firing flares for the fw190 spawning mission easier as well. I'll give it a try with just the mission file, it very likely behaves the same way as in SP; I just wasn't aware of this possibility and thought the game needed the .msnbin always, thanks.

Edited November 18, 2022 by firdimigdi

[firdimigdi]

On 11/18/2022 at 10:48 AM, chiliwili69 said:

If you are going to recreate the mission you can use the I-16 plane, so we don´t need to open the cabin to see the stutters better. Also, the I-16 is the best plane to do quick roll.

 

Actually turns out I had it so you can choose whatever allied plane you want and the I-16 was already available.

But here's all the test missions again without the msnbin for compatibility.

These have to be unzipped in data\Multiplayer\Dogfight

 

 

stuttertest - two missions vehicles vs planes.zip stuttertest - plane spawner on flare.zip

[chiliwili69]

On 11/19/2022 at 5:55 PM, firdimigdi said:

the I-16 was already available.

 

Yesterday I was testing those scenarios, I think the best one is the plane spawner on flare. 

I didn´t know how to run it in MP but I saw that I had to config a new server with those missions and also allow infinite anmo to have infinite flares. The I-16 type is ideal for this test.

 

With just one plane no stutters, but with 20 planes the stutters are noticeable. Then with 30 is worse and with 50 planes is really dizzing. All running at 80fps always.

It is only seen in all objects outside the plane, specially landscape.

[IckyATLAS]

27 minutes ago, chiliwili69 said:

 

Yesterday I was testing those scenarios, I think the best one is the plane spawner on flare. 

I didn´t know how to run it in MP but I saw that I had to config a new server with those missions and also allow infinite anmo to have infinite flares. The I-16 type is ideal for this test.

 

With just one plane no stutters, but with 20 planes the stutters are noticeable. Then with 30 is worse and with 50 planes is really dizzing. All running at 80fps always.

It is only seen in all objects outside the plane, specially landscape.

One test you should do for one, 20, 50 planes is to look from outside the plane from a fixed camera (F11) and then look around and see if you have the same level of stuttering.

I noticed that if you are looking around from an outside the plane camera in flight (F2) then you may have more visible stutter than with a fixed camera in the same configuration. This probably due to the fact that the camera is being connected to the plane and is dependent on plane motion in the environment.

[kissTheSky]

On 7/6/2022 at 5:43 AM, Regingrave said:

That can be very useful, thank you for your research!  We'll take it into consideration, but Scaleform is a quite heavy framework as it is, this is a longstanding limiting factor for us, as we can't really do anything about it.

@Regingrave, has there been any further development into this? Especially with Normandy, game is very distracting to be immersive. 

[pellson]

Did you find a solution for this?

The game runs buttery smooth for me at 30 FPS in VR with Reverb G2 on quick mission (with motion reprojection of course since I have a shitty 2070s, therefore locked to 30, I can live with the artifacts), but as soon as I try the campaign with many AI units the game is just a stuttery mess, EVEN THO I still have full 30 FPS alll the time just like in quick mission. Graphics settings doesn't matter, have tried everything on Low with 50% resolution.

 

 

[Two_Stroke]

I’ve been troubleshooting this for over 5 months. Please let me know if you discovery anything new. I want to love this game so much, but the stutter is extremely disheartening. 

[PeYBoL]

Since last pach, I also have the landscape stutter issue. It only happens with cockpit camera. With outside cameras it looks ok with no landscape stuttering.

Edited January 22, 2023 by PeYBoL

[PA_Willy]

Reduce the refresh rate you normally use by one FPS and limit the maximum FPS your device will render to that value. This is done from the Nvidia control panel.

For example, if your device renders at 60FPS, limit that factor to 59 and see if that improves your microstuttering.

 

[OG_NickNack68]

Greetings All,

Since the FC 2 update last week my smooth VR experience has become stuttery as described in the above posts. I don't have the technical expertise evidenced by your posts above but I noticed the same issues with changing graphics settings, resolutions, etc. Doesn't seem to change the issue. Just found this post this morning as I was searching for anyone having issues since the update. I will try to investigate by returning to monitor play and see if I have the same issues. Somewhat disheartening after just getting VR last month and finally getting it set to run smooth and then the update and now it's very distracting. I was so excited about the career mode for FC and the added immersion with VR. I'll check back here later and add to the investigation if I can.

Nick

[savagebeest]

On 1/22/2023 at 11:56 AM, PA_Willy said:

Reduce the refresh rate you normally use by one FPS and limit the maximum FPS your device will render to that value. This is done from the Nvidia control panel.

For example, if your device renders at 60FPS, limit that factor to 59 and see if that improves your microstuttering.

 

I will try this. This horizon micro stutter is killing the experience for me on vr. I don’t want to have to uninstall this game as I do enjoy but when it used to run properly. I have faith in 1C they always deliver. 

[savagebeest]

Sadly setting gps to minus 1 fps lower (71 in my case) didn’t have any affect. However, I did set all graphics to max and I mean everything to its limit (no supersample) and the horizon micro stutter is nearly gone to the point where it’s playable for me and I’m super sensitive to frame rate. 
 

i7 12700kf

4070ti 12gb

3200 16 ram

dpvr e4 steam headset (opencomposite)