investigating Large performance drop in VR after updating to 2.8

[darkman222]

@BIGNEWY Here is some more testing data. With my Varjo Aero this time too.

Seems like an additional 1 ms CPU frame time happens using the latest OB compared to stable with both VR headsets. Creating the negative spikes in the GPU graph of the Aero.

I have the same 1 ms CPU frame time loss on the Pimax too, although the CPU graph shows that the CPU frame time for the target frame rate of 90 fps should be sufficient, the GPU frame graph shows the loss of 1 ms. CPU graph shows it too, but is not marked as the bottle neck, but it could be too for the Pimax.

So maybe it was misleading to me, that the performance drop using the Pimax came from the GPU , it might be coming from the CPU although the graph shows that the CPU frame time still should be sufficient.

 

 

 

 

 

EDIT: In my ealier post with OB 2.8.0.32235.1 Open Beta the loss was only 0.3 ms on the CPU but about 2 ms for the GPU. Just as a heads up. Direct link to post below:

EDIT2: I also attached my test mission with just an empty mission only 1 F16 over Batumi. Thats the same mission I used in my original test above.

perf_test_caucasus.miz

 

System Spcs:

Edited November 30, 2022 by darkman222

[YoYo]

Hi, can you campare Shadows = Medium and Terrain Object Shadows = Flat Only? Stable vs OB. The Bigest issue is Flat Only in 2.8 but in TOS setting according me. 

[BIGNEWY]

8 hours ago, darkman222 said:

@BIGNEWY Here is some more testing data. With my Varjo Aero this time too.

Seems like an additional 1 ms CPU frame time happens using the latest OB compared to stable with both VR headsets. Creating the negative spikes in the GPU graph of the Aero.

I have the same 1 ms CPU frame time loss on the Pimax too, although the CPU graph shows that the CPU frame time for the target frame rate of 90 fps should be sufficient, the GPU frame graph shows the loss of 1 ms. CPU graph shows it too, but is not marked as the bottle neck, but it could be too for the Pimax.

So maybe it was misleading to me, that the performance drop using the Pimax came from the GPU , it might be coming from the CPU although the graph shows that the CPU frame time still should be sufficient.

 

 

 

 

 

 

EDIT: In my ealier post with OB 2.8.0.32235.1 Open Beta the loss was only 0.3 ms on the CPU but about 2 ms for the GPU. Just as a heads up. Direct link to post below:

 

 

thanks for sharing 

[Jenson]

On 11/27/2022 at 7:26 PM, Baldrick33 said:

I think the Open Beta concept is quite sound. People get to choose if they want to be early adopters of the shiny new stuff, with the knowledge there are risks some stuff might get broken. This isn't some kind of corporate app which undergoes massively time consuming regression testing to ensure there isn't a bug in the reporting output (with a standard build client PC), the users provide feedback across a myriad of hardware and volume of testing which a smaller closed group could never achieve.

Personally I would prefer a more conservative approach to promoting to stable, such that stable is a more bug free environment but I guess ED know that the majority would prefer not to wait too long for new modules and updates.

The multiplayer aspect is a massive conflict as DCS is not predominantly a multiplayer game and Open Beta will often create issues for certain groups of players, be it hardware or features installed on specific aircraft which either creates bugs or imbalance. I guess there is some history as to why Open Beta is widely adopted for multiplayer but with a clean sheet of paper it seems an illogical choice if the aim is for stability and fairness.

Agreed. 

But with Steam edition, there is only two version available to choose, 2.5.xxx or Open Beta, I hope ED could let us choose the latest stable version as well. 

[BeerNfrites]

Has it been acknowledged that there is a problem that needs addressing or are we still at the stage of gathering data?

I keep seeing statements that not everyone has seen performance drops and feedback has been mixed.

I feel personally like the blame is on me, it’s my fault my performance has gone down, rather than there being an issue with the release that needs fixed. 

[darkman222]

If you have the hard disk space, install the stable version before it also gets pushed to 2.8.

You can blame 2.8 for the performance loss for sure.

My personal guess is that people who dont see a performance drop just have not a system that was on the limit before 2.8. So they just dont realize it.

If people in VR dont see a performance drop they might be using some kind of motion smoothing which does not care if they had 65 fps but now have only 55 fps, because it clamps it at 45 fps anyway. Just wild guesses.

This needs to be fixed or adressed or whatever. I dont want to accept a performance drop and rely on that we get multithreading soon or not so soon. I am glad that it was announced. But it was announced for next year. That means 13 months from now.

Edited December 10, 2022 by darkman222

[YoYo]

5 hours ago, darkman222 said:

If you have the hard disk space, install the stable version before it also gets pushed to 2.8.

You can blame 2.8 for the performance loss for sure.

My personal guess is that people who dont see a performance drop just have not a system that was on the limit before 2.8. So they just dont realize it.

If people in VR dont see a performance drop they might be using some kind of motion smoothing which does not care if they had 65 fps but now have only 55 fps, because it clamps it at 45 fps anyway. Just wild guesses.

This needs to be fixed or adressed or whatever. I dont want to accept a performance drop and rely on that we get multithreading soon or not so soon. I am glad that it was announced. But it was announced for next year. That means 13 months from now.

 

You have right, however if your motion reprojection (Steam VR) is set to 45 (half of refresh rate) its visible too, because DCS cant hold 45 frames all the time like before (you see jumping of FPS between 40 to 50) 2.8 with Shadows Flat on (for example for me). If the user use WMR mod or has motion smoothnes on the level 30 fps perhaps he doesnt see it so quickly but its present still.

[muzica9]

22 hours ago, darkman222 said:

This needs to be fixed or adressed or whatever. I dont want to accept a performance drop and rely on that we get multithreading soon or not so soon. I am glad that it was announced. But it was announced for next year. That means 13 months from now.

 

i can bet now that it will be in q4 2023 or q2 2024

Edited December 10, 2022 by muzica9

[YoYo]

Some checked after the Fridays patch? Can we use *Flat on” shadows again like in 2.7?

Now PG map is affected? 

 

[zildac]

1 hour ago, YoYo said:

Some checked after the Fridays patch? Can we use *Flat on” shadows again like in 2.7?

Now PG map is affected? 

 

For me, no. Terrain shadows Flat still results in around a 20FPS loss in VR. So I'm staying with Off for the time being.

[speed-of-heat]

ditto

and to be fair there wasn't anything in the patch notes to say they had fixed it 

 

[freehand]

Flat Shadows.

I done a quick pancake test not tried in vr but cockpit I lose 20 ish but external on the F18 no drop at all.

Also the FPS in the cockpit are the same with shadows Hight or Low no difference but still lose the same 20 ish frames. 

. 

Edited December 17, 2022 by freehand

[lagner]

5600x, 3080ti, 32GB RAM, Oculus Rift S.
One and the same stage and military equipment:
2.7 - 90% usage GPU;
2.8.1 - 100% GPU usage and FPS drops;

Empirically, it should be assumed that my system lost about 20% of performance.
This was the last straw.
ED won't get my money anymore.

UPD: 

Same track, same settings, no VR launch:
2.7 - 233 FPS;
2.8.1 - 165 FPS;
The difference is more than 40%. Great weekend (NO) with Black Shark 3.

Edited December 17, 2022 by lagner

[Flyingfish]

I've updated to Open Beta 2.8.1.34437 and my performance is much worse in VR than it was in the previous open beta 

[Andrei]

VR Tests using plasma's map benchmark

DCS 2.8.0.33006
Average FPS: 42
Average CPU FT: 12.6 
Average GPU FT: 23.2

DCS 2.8.1.34437
Average FPS: 38
Average CPU FT: 14.9
Average GPU FT: 25.2

Frametimes - lower is better. So 2.8.1 seems worse in terms of performance.

There can be other variables at play since I got Win11 H2 update and newer nVidia drivers, but the issue with GeForce cards on Win11 has been fixed.

[C3PO]

Just leaving this here.

 

 

[muzica9]

7 hours ago, Andrei said:

VR Tests using plasma's map benchmark

DCS 2.8.0.33006
Average FPS: 42
Average CPU FT: 12.6 
Average GPU FT: 23.2

DCS 2.8.1.34437
Average FPS: 38
Average CPU FT: 14.9
Average GPU FT: 25.2

Frametimes - lower is better. So 2.8.1 seems worse in terms of performance.

That is another 10% loss. But hey!!! You got new lights on PG map! 

[WipeUout]

I re-installed Stable 2.7.18 as I did not have the 4080 then and wanted to update my baseline results.

Here are my results with HIGH preset setting using my test track file (F-18, Caucasus):

                      2.7.18                2.8.1          Delta

GPU FT          13.4                 14.5            +1.1ms

CPU FT          10.2                 12.7            +2.5ms

Seems that the CPU is taking a massive hit.  This could explain why some are seeing huge FPS drop if they were already at the edge of their refresh rate with 2.7.18.  Results with previous version 2.8.0.33006 are the same as latest 2.8.1.  I use the following mods: VR Shader Mod, ReshadeVR_sharpen_color 4.9.1, F-14_VUX_v2.2, Taz1004_Better_Trees_v5_Med_Low, VR Loading Screen fix and PimaxXR/OpenComposite0.6.3/OpenXRTK.  No plan to go back to OB until the problem is fixed.  Unfortunately, only one version is possible with steam.

 

 

[muzica9]

6 hours ago, WipeUout said:

 No plan to go back to OB until the problem is fixed.  

 

 

Will never be fixed. New clouds, new map light, etc = New DCS. Every major update was a perfomance killer and same way this time.

[WipeUout]

Do not agree.  There is something causing this drop and not everyone is experiencing it, they just need to put the finger on it and it will be fixed eventually.  If it was just added processing then everyone would see a drop, which is not the case.  ED would not waste time investigating and asking for data! 

Edited December 19, 2022 by WipeUout

[some1]

For a start, Flat Terrain Objects Shadows are still broken and cost much more FPS than they used to in 2.7. 

Although 'high' DCS preset is not using those IIRC.

Edited December 19, 2022 by some1

[darkman222]

49 minutes ago, WipeUout said:

Do not agree.  There is something causing this drop and not everyone is experiencing it, they just need to put the finger on it and it will be fixed eventually.  If it was just added processing then everyone would see a drop, which is not the case.  ED would not waste time investigating and asking for data! 

 

Not sure if it is just that not everyone cares and complains because his PC has not been on the limit before the update, thus this person just does not realize it.

[Marklar]

56 minutes ago, WipeUout said:

Do not agree.  There is something causing this drop and not everyone is experiencing it, they just need to put the finger on it and it will be fixed eventually.  If it was just added processing then everyone would see a drop, which is not the case.  ED would not waste time investigating and asking for data! 

 

Don't hold your breath. 

 

[lagner]

 

9 часов назад, Marklar сказал:

Не задерживайте дыхание. 

 

Great. The 2.8 stable version could be the starting point for selling aftermarket flight accessories.

Edited December 20, 2022 by lagner

[WipeUout]

12 hours ago, Marklar said:

Don't hold your breath. 

What a disappointing comment from NineLine!  Hope this is not the official ED stand on the issue.