Desktop frame rate and vsync effects on FPS

[mark007777]

Hi,

This evening I was reading various forum posts both old and new, talking about both vsync on vs off as well as setting the desktop resolution/hz to certain lower values to improve msfs performance. (I would try anything at this point to get an extra frame or two )

I was surprised with the results however, and I’m not sure what to think. I would love your opinions on both of my findings below. Here are my specs and findings anyways.

TV: LG C1 77" (Set to 4k @ 120hz with g-sync on, v-sync on for most games)
GPU: MSI RTX 3080 stock
CPU: AMD Ryzen 9 5950x stock
Headset: Reverb G2

So I performed the following two tests.

1. I set the desktop to really low resolutions / low hz. None of these improved the frame rate according to the in game dev fps counter.

Setting the hz of the screen to say 24hz, ** limited the rendered frame rate of the game to 24hz. Is this right? ** This means the frame rate of the screen actually is in some way tied to the frame rate of the display which in some way worries me, that we do need to be really careful about what hz we set our displays to. Should we sync our display and headset hz if possible? I tried setting the TV to a custom resolution of 45 and 90 but no major difference. Perhaps other hz in the 30 range is also a valid hz for when vsync is on?

My question here is, how are some people seeing that setting the display to 24hz or lower resolutions, is getting them higher frame rates, say 30/40+ in game? Is this gpu / gpu setting dependent?

2. For vsync on/off, I did notice in various test runs over the past few months, that for me vsync off and HAGS on, gives me the best fps, but just barely, by 3 or 4 fps so I leave it like that. So my nvidia settings default to g-sync on and vsync on for all other games, but vsync off specifically for msfs.

While testing this, I however noticed, in game, my frame rate might show say 37fps in NYC for example (I purposely set render resolution very low by the way to notice differences more) but when I press the start button to bring up say the nvidia control panel to mess with resolutions/hz, i saw in the background, the frame rate of the game jump to 42/43 fps. This is around a 15% improvement! Switching focus back returns the frame rate to the same, 37fps. (This is with the escape / msfs menu open so the plane is not moving / the scene is not changing).

I am trying to explain this and I think it something to do with vsync. Is it that when the msfs window is in focus, vsync is enabled (because its vr and something forces vsync to be on?, even though I force it to off in the nvcp?) and the frame rate drops (surprisingly not to 90, 45, or 22.5 but something randomly in between…like 37 ) and when I switch focus out to some other window / start menu, vsync goes back off (I guess as its flight sim still active but maybe some vr mode becomes inactive?? turning off the forced vsync). I’m so confused to be honest

It leads me to a few questions, are others seeing this? Is there some 15% performance gain to be had here if we make sure some other window is active while playing msfs in vr, or if we can somehow force vsync off in game. My understanding with vsync is that if fps is so low (20s/30s), tearing shouldn’t be noticeable? Maybe that’s not true for VR?

All opinions welcome. I truly hope this is some free fps we can all gain?

Cheers.

[Majak2644]

GPU: AMD RX 6900 XT
CPU: AMD Ryzen 9 5900x
Headset: Reverb G2
Windows 10 21H1

There are quite many people observing better performance (for months or even years) when the game is out-of-focus. However, this seems to be not limited to MSFS only; I do experience the same with DCS for example.

My GPU is AMD, yours is nvidia. So I guess it’s not related to the GPU manufacturer.

Maybe it’s a Reverb G2, OpenXR or Windows Scheduler issue.

Some people have reported that enabling optimiziation for Background Services helps, others reported that giving the game only a few CPU cores and/or making the game process Low Priority helps. But not for me.

[Mayhem6633]

Lots of people say lots of things. I just wish people would back up their claims with actual testing. I think most of these claims are a placebo effect. Everyone should be using a program like CapFrameX to measure performance gains before reporting them. CapFrameX measures in-game performance frame-by-frame and reports the average framerate but more importantly reports the 1% and 0.1% lows.

For example, here’s a measurement I made last night. Notice my average framerate is only 30.1 fps but also notice my 0.1% low is above 20. This flight was a completely stutter-free and smooth flight over a large city with photogrammetry. I’m finding “above 20” for the 0.1% lows is the key to a smooth experience. I’d welcome any tips that can show improvements to the 1% and 0.1% low framerates.

[Majak2644]

@Mayhem6633 Not sure if you meant me by “Lots of peope say lots of things.” or if you meant the guys claiming that doing silly placebo things enhances framerate and frametimes.

Here’s a “proof” for the out-of-focus vs in-focus observation which I analyzed 3-4 weeks back on my machine when I noticed it (also giving silly things a try). I must admit that it’s only a single 20s CapFrameX run using the featured New York flight, looking to the left at the buildings below at the very beginning of the flight, but it should showcase that the out-of-focus vs in-focus issue exists on some machines. At least it existed 3 weeks back on my machine in MSFS (and really noticable when motion reprojection was acive, thouth I usually play without it); I didn’t play MSFS since then, but I am noticing that issue with DCS which I am currently playing; in DCS with my current settings being out-of-focus raises FPS from say 60 fps to 82 or so; however in DCS I could imagine that it’s the user-input (stick/hotas/pedals) which causes it. I mean, it could be that when out-of-focus the game simply does ignore user input (on the other hand, I can observe the 60 vs 82 fps difference also when watching replays, where user input should not be a limiting factor).

[Mayhem6633]

@Majak2644

Not sure if you meant me by “Lots of peope say lots of things.”

No. It was a general comment that it’s difficult to know which suggestions to believe and which not. I know lots of the claims are true but many more are not so it gets frustrating sometimes. I made that post to let people know there’s an actual tool they could be using, like you have.

But back to your data. Thanks. It’s clear there’s much less consistency in the “in focus” data so there is something to this. Like I was showing in my charts, it’s not about the framerates. It’s about the consistency in the framerates. Here’s my time diffs and L-shape charts for the data I posted above. My frametimes are pretty consistent thus I get smooth flights.

Let me posit a thought on what is happening. I don’t use motion reprojection either. If you notice on my chart I’m showing those measurements as “30fps capped”. I started capping my framerate because I was seeing what you showed, lots of variations between the frametimes. I also noticed that my GPU was maxed out at 100% usage. I think when the GPU is maxed it has a hard time giving consistent performance. When I cap my framerate my GPU doesn’t run at 100% utilization. It’s 80-90%. That’s when my 1% and 0.1% lows improve so I think because the GPU isn’t maxed that it has some headroom to perform better for the low framerate times.

So maybe it’s the same thing with this situation. People are running uncapped. So when the game window is in focus Windows gives the game a hgih priority and that pushes the GPU to the max leading to inconsistent performance. On the opposite side, when the game is not in focus, the game doesn’t receive the highest priority from Windows and therefore the GPU works a little easier and therefore the consistency between frames is better. It might be a logical conclusion.

But in the end and after comparing my data with your data I think you’re chasing a pipe dream. I think the window being in or out of focus is of little consequence. There is an improvement but look at my time diff chart compared to yours. My blue color extends to around 87% of the time versus your graphs that extend to 17% in the best case. I think running the window out-of-focus makes very little difference. This is why I like seeing data. Yes, you did find something but I think it’s of little consequence.

[Majak2644]

Agree @Mayhem6633, I also prefer to play capped for that reason (I use lower quality such that my cap can be somewhere between 40-45 rather than 31 fps; need to play more with it to find a sweet spot for me; 31 fps is consistent but it feels a little bit too choppy for me).

I cannot prove this now, but back then when I tested with motion reprojection I had set the cap to 31 fps via Adrenalin driver. And in that case the in-focus MSFS was stuttering like hell (it showed ~22.5 fps in the OpenXR overlay reprojected to ~80 fps) whereas the no-focus MSFS could hold its ~30 fps reprojected to 90 fps which felt very smooth. So my impression was similar to what @mark007777 observed: It felt like some VSYNC was active when the game was in-focus. Similar thing I do observe with DCS: With in-focus its exactly 60 fps fixed in my test scene, not more, not less. With no-focus it’s 74 (just measured it; my 60 → 82 from above were little bit exaggerated…). If I set higher quality in DCS, the in-focus DCS drops from exactly 60 to exactly 45. Could be coincidence and I need to test more with other settings, but right now it looks like some active VSYNC in the in-focus case.

[Mayhem6633]

@Majak2644

back then when I tested with motion reprojection

I get it. My experience is strictly valid when motion reprojection is disabled. When you turn it on then the framerates get locked to only those preset rates as you’re explaining. So I can see where even a minor improvement might make the framerate with motion reprojection on jump to the next higher value and thus you experience much improved performance.

I leave MR off in both games because I’ve been able to find smooth game-play as I described but I understand that 30 fps might not be good for some people. That value is for me and you need to find what’s best for you. The major point is the framerate consistency. That’s what motion reprojection does. It fills in frames so you get better consistency.

Unfortunately for me I get bad artifacting with it on, moreso in MSFS than DCS, but if I can get satisfactory performace with it off then I do that. And luckily I’ve been able to get great performance with it off. In MSFS I’m running with 85% resolution in OpenXR and 100% in-game. So I’m getting a great picture on top of the great performance.

[mark007777]

Thanks for the replies guys, and wow great graphs from that capframeX. I must give it a try. Are we still however theorizing as to what causes the in-focus vs not in-focus max frame rate differences? Is there some way to proove/disproove why that might be happening.

I’d be interested to see is there a best of both worlds here, capping the frame rate, but also going out of focus to see if it helps average frame times also go up? ie best of both worlds?

Do you see any benefit on your capframeX with the out of focus scenario @Mayhem6633 ? Does it improve your graphs even further?

[Majak2644]

Yes, @mark007777, we are still guessing. :-/

Would be interesting get more people experiencing this to figure out if it is related to some hardware/software combos. So far with our systems I see that we both are on AMD CPU’s (I’m on X570 chipset) and both have the G2.

P. S.
When I tried yesterday evening to reproduce this in MSFS, I failed. But that doesn’t mean that the issue is solved for me since it was the same choppy experience with both in-focus and no-focus. Not sure if it’s due to some software changes on my system (it’s 3 weeks back that I played MSFS and meanwhile I updated GPU drivers + re-installed WMR/OpenXR and probably some Windows Updates have been installed) or if it’s just some randomness (e. g. maybe Windows Scheduler decided to give other priorities yesterday). In DCS it was still reproducible though (with and without motion reprojection enabled).

[DerKlausi]

Thanks a lot for substantiating the discussion with actual numbers!

I’ve compared MSFS window in focus vs. Windows start menu focus myself, see the results below. I’ve done two runs with each setting with MSFS restart after each run.The results of runs with the same setting agree by 0.3 FPS so I included only one each.
I’m losing a little in average FPS but gaining almost 4 FPS in min FPS. Frame times are a tad higher with MSFS out of focus, but IMHO 0.5ms is within margin of error.
For my full specs and settings as well as benchmark procedure please see

Standard Bench Flight for Performance Comparison

One of the main issues with comparing all the user feedback in terms of performance issues and regressions is the flight conditions used. Depending on whether you’re flying low over photogrammetric cities, dense forests or 2k up in the air over plains will easily make 10…15 FPS difference. Different airplanes and weather conditions as well as lighting conditions (time of day, season) add to that. That’s why I’m doing my performance analysis using the Flight Recorder addon with a pre-recorded fli…

[Majak2644]

Though I used different quality settings, I have similar results on my system with your recorded flight from the other thread @DerKlausi.

[CaptainNuts2000]

@Majak2644

Maybe it’s a Reverb G2, OpenXR or Windows Scheduler issue.

Possibly pointing to Scheduler: First off, this is not the case (for me) on MSFS currently (with one exception, see below). But I know the effect (and the disbelief) from X-Plane, where it is very pronounced (forgot the numbers but the difference is big enough to decide between “no fun” and “playable” in dense areas) and there I had it first on the Rift S, then on the Reverb G1 - so the HMD wasn’t the culprit there.

Secondly, I get this effect in MSFS when my Reverb went to sleep and I reactivate it, then it sometimes ends up in some kind of upset state and the reduced, stuttery framerate recovers when I Alt+Tab MSFS out of focus. That sounds a lot like something ended up on the wrong CPU core.

[Majak2644]

Fun fact: I observe the complete opposite if I work with really low resolutions. Then in-focus becomes much better than no-focus.

[DerKlausi]

I’ve done some more testing using the latest NV496.84 driver which itself did not lead to any significant performance differences vs. 496.49 for me. However, what I did was to compare the following window focus situations:

1. MSFS is the only window visible on-screen and in-focus (mouse-clicked)
2. MSFS plus Start Menu raised via Windows key
3. All windows incl. MSFS minimized so that only the Windows Desktop is visible.

I’ve repeated the tests 3 times each but variations were small and results reproducible so I included only one plot each. The results show that with my setup I’m gaining even a few more minimum FPS (and +1FPS average FWIW) by reducing 2D visible output to Desktop only.

What’s more, when checking Windows Task Manager, ANY additional windows visible on-screen (WMR App, Flight Recorder, …) will lead to a small additional GPU load of 1…3% with my RTX3090. Maybe this is more significant for users with lesser GPUs so hopefully this info will be helpful for others.

Now let’s hope SU7 won’t bring us any performance regressions…

[N3804Q]

I have the same problem with my Pimax 8k plus using windows 10. My msfs is coming from steam. If we can project msfs inside the steam vr window, we may can fix the problem. Msfs I. One windows and steam in another is penalizing our VR performance. Any expert programming? Or creating some mod?