reported FPS Lost after DCS 2.7.7.14727 patch

[Hiob]

1 minute ago, Miro said:

In my case, the graphics card clogs up, wear reaches 100% and frames drop down. In previous patches (the working ones) it didn't get clogged. If the DCS had changed graphically, it would have been understandable, but nothing has changed, except for water, the reduction of which to low does almost nothing.

Never claimed, that the idea, is A or even THE ultimate solution. Just another try to help find the root of the problem.

There can always be several bottlenecks.

[Limaro]

Could please someone who has absolutely no FPS loss post some data, like dxdiag, graphical settings, trackfiles - so that we have something to compare with?

[Lange_666]

2 hours ago, Hiob said:

One thing to notice in the provided screenshots by Lange_666 is that there is obviously a difference in core usage of the cpu. May it be, that there has been updates/changes introduced (either by microsoft or by the game core) regarding core-scheduling? As far as I learned, the approach to core/thread-handling in win11 is supposed to be the culprit for the sub par performance of certain cpus. May it be, that a similar (or even connected) problem leads to the issue at hand?

Underusage of GPU and seemingly more load on cpu cores seem to a repeating pattern....

Just ran a CPU core test, it jumps between each of them, stable and open beta. I still run Win 10 (no joy in upgrading to Win 11 with my system, i don't pass the tests).

Mission: Instant action: F-16C: Free Flight: Caucasus, exact same settings in both.
(VSync + FPS limiter off which i normally have on and set to 60  to let the FPS "fly" un-restrictedly).
Waited until the mission text in the upper right corner vanished. Then took 3 shots within the minute, +/- around the same time in the mission, same terrain same clouds.

Stable 2.7.6.:

    

 

Open Beta 2.7.7.

  

Edited November 4, 2021 by Lange_666

[Flia]

This looks like a really serious problem in DCS. 20 percent difference is quite a lot.

[Hiob]

42 minutes ago, Limaro said:

Could please someone who has absolutely no FPS loss post some data, like dxdiag, graphical settings, trackfiles - so that we have something to compare with?

I did, here - 4K, custom high settings, in Marianas, Trackfile included. No diag-files, though. The result was, exactly the same performance as before the patch. I doubt, that it is of real benefit though. My Specs are Ryzen 5900x, 64 GB 3600 Ram, Gigabyte RTX3080. I will try to provide the diag files, when I find the time.

https://forums.eagle.ru/topic/269510-dcs-27-detailed-4k-benchmark-with-settings-new-scene-for-marianas/page/2/#comment-4807505

 

@Lange_666

Could you try to force DCS to (two) certain cores (on a single die, if you have a multi-die cpu) and see if it gives you any results?

Edit: Just seen your Specs - I have to admit, that this particular point does apply more to multi-die ryzen cpus. Really don't know if it makes ANY difference for an i7. But no harm in trying.

Edited November 4, 2021 by Hiob

[DmitriKozlowsky]

As I understand it. There are two types of cores in Intel. Performance core and efficiency core. Performance cores are 0,2,4, etc.. Efficiency cores are 1,3,5,etc.. 

Limit DCS to performance cores, and see if that improves. But I doubt it. I don't think DCS is multithreaded. Most of FPS comes from GPU.

 

[Hiob]

46 minutes ago, DmitriKozlowsky said:

As I understand it. There are two types of cores in Intel. Performance core and efficiency core. Performance cores are 0,2,4, etc.. Efficiency cores are 1,3,5,etc.. 

Limit DCS to performance cores, and see if that improves. But I doubt it. I don't think DCS is multithreaded. Most of FPS comes from GPU.

 

No, I think that only applies to the latest (12th) generation. In previous Intel generations all cores should be equal. You need to differentiate between physical cores and threads, though.

The task manager will show 8 cpus for a 4-core cpu. Usually the even numbers (0,2,4 etc) are the physical cores and the odd numbers are only logical threads.

(The underlaying tech is "hyperthreading" (smt), introduced with Pentium 4 and basically standard by today)

Edited November 4, 2021 by Hiob

[DmitriKozlowsky]

45 minutes ago, Hiob said:

No, I think that only applies to the latest (12th) generation. In previous Intel generations all cores should be equal. You need to differentiate between physical cores and threads, though.

The task manager will show 8 cpus for a 4-core cpu. Usually the even numbers (0,2,4 etc) are the physical cores and the odd numbers are only logical threads.

(The underlaying tech is "hyperthreading" (smt), introduced with Pentium 4 and basically standard by today)

 

You are right. I meant to say Performance Core and Efficiency THread. Core is physical and is assigned to threads 0,2,4, etc..  Why efficiency threads are 1,3,5,etc..

 

[Lange_666]

3 hours ago, Hiob said:

@Lange_666

Could you try to force DCS to (two) certain cores (on a single die, if you have a multi-die cpu) and see if it gives you any results?

Edit: Just seen your Specs - I have to admit, that this particular point does apply more to multi-die ryzen cpus. Really don't know if it makes ANY difference for an i7. But no harm in trying.

Result is the same, just more CPU usage on the cores used and less on the not used cores.
Feels overall a bit less smooth, maybe because the cores now run close to max usage (on some moments they went op to 98-99%).

OB 2.7.7:

All cores (Hyperthreading OFF)

 

2 cores (0 and 2 in the shot below = CPU 1 and 3 but test with cores 1 and 3 (CPU 2 and 4) was the same), Hyperthreading OFF

 

Result with Stable 2.7.6. was also the same, all or only 2 cores same result, but better FPS compared to OB (see post a bit up):

[Hiob]

I still have my old rig standing by - it's quite similar to yours (4790K, 1080ti, 16GB Ram).

If it's still running, I could try to reproduce your results. Could be interesting.

[felipesm]

I am starting to believe that you(ED team)are being negligent about this issue. It is not possible didn't have identified the problem until now. (After two weeks and two updates). 

Edited November 4, 2021 by felipesm

[Enduro14]

So I Just did the F16 free flight cauc mission as well, and watched the cpu usage, now I don't have any way to compare to previous but I am seeing the cpu up to 80% 90% and some times more, then it drops to literally no usage.  I used Process lasso to set dcs to all physical cores on my I7 8700k that is overclocked to 5.0ghz.

 

Took screens but the overlay from msi afterburner didnt show on screens of course...  

 

Also to note 1080ti pegged at 99% the whole time granted did it in 2d on my 4k screen.

 

Just a side thought, but as the multi threading was in news recently and most stuff done for it, I wonder if some elements are now Multithreading from these recent updates?  Saw high usage spread over all physical cores in above test.

Edited November 4, 2021 by Enduro14

[Hiob]

15 minutes ago, felipesm said:

I am starting to believe that you(ED team)are being negligent about this issue. It is not possible didn't have identified the problem until now. (After two weeks and two updates). 

 

Oh lord, really? 13 Posts in this Forum and you're already spreading senseless allegations? What is this supposed to achieve, other than spreading a bad mood?

I'm so sick of complainers like you!

[Sr.]

15 minutes ago, Enduro14 said:

So I Just did the F16 free flight cauc mission as well, and watched the cpu usage, now I don't have any way to compare to previous but I am seeing the cpu up to 80% 90% and some times more, then it drops to literally no usage.  I used Process lasso to set dcs to all physical cores on my I7 8700k that is overclocked to 5.0ghz.

 

Took screens but the overlay from msi afterburner didnt show on screens of course...  

 

Also to note 1080ti pegged at 99% the whole time granted did it in 2d on my 4k screen.

 

Just a side thought, but as the multi threading was in news recently and most stuff done for it, I wonder if some elements are now Multithreading from these recent updates?  Saw high usage spread over all physical cores in above test.

 

From my experience, that is high on the list of absolute worst map/missions to fly. I think it's mostly the application of whatever cloud preset it uses.

Worst jaggedy edged, bouncy clouds in DCS.

[Lange_666]

But for a test like this it doesn't matter what mission/map/airplane is used.
I'm seeing the same thing with other aircraft/maps/missions only with (obvious) different FPS values.

[Enduro14]

Followed this video and i have to do more testing and have to check vr yet but in the F16 mission above in 2d just staying static I just saw an increase of 10-15 fps.  But more testing is needed to verify

[Hiob]

The number of svchost.exe's is just showing how many background processes are running. I'm not saying, that the above approach isn't beneficial, but you should de-bloat your OS anyway to optimize performance.

[Enduro14]

3 minutes ago, Hiob said:

The number of svchost.exe's is just showing how many background processes are running. I'm not saying, that the above approach isn't beneficial, but you should de-bloat your OS anyway to optimize performance.

True its certainly debloated but seemed like an improvement but vr is still not there pre patch neither is 2d.  

[Hiob]

Just now, Enduro14 said:

True its certainly debloated but seemed like an improvement but vr is still not there pre patch neither is 2d.  

10-15 FPS is a lot though!

[peachmonkey]

here's a brief explanation to that registry setting:  SvcHostSplitThresholdInKB

It also has the table of values per the RAM installed.

If your system is already debloated and all unnecessary services have been stopped this won't help in your pursuit of higher frames.

https://www.tenforums.com/tutorials/94628-change-split-threshold-svchost-exe-windows-10-a.html

[Enduro14]

Ill keep it brief mode,  Gpu wasn't boosting all the way to its full clock,  when I enabled performance mode in process lasso then card boosted up.  Probaly what got me more fps in that one scenario.   

What i am seeing without a doubt is much higher CPU usage on all cores I have set for dcs.

 

still have to have the dcs pd dropped in order to get close to pre patch fps.  Where pre patch would not have dropped pd and also had settings higher.  Guess ill just check back after future updates and see if anything improves.  I'm burnt out trying to troubleshoot this.

[Funflite]

52 minutes ago, Sr. said:

From my experience, that is high on the list of absolute worst map/missions to fly. I think it's mostly the application of whatever cloud preset it uses.

Worst jaggedy edged, bouncy clouds in DCS.

I'm pretty sure in that the clouds systems in the DCS 2.7 version is the beginning of the loss of frame rate in VR.It's really nice in 2D and this why i'm not flying in VR mode because a this massive frame drop in VR at least half of the 2D mode.Clouds in very angry VRAM consummer in VR  a maybe ED will make an option a disable clouds in the VR menu option at least in Open beta version and see if VR users see a regain in fluid flying in MP or others missions or campaigns.I'm not  sure that we on ourselves to find a solution on this concern but by ED.I make a test by editiong a simple mission with and without clouds...it's amazing on how clouds is consumming ressources.The clouds is a fine things but by the way it is manage in DCS is not good in VR.

[pablofe]

I will do more detailed data collection on my end, but I wanted to check whether all of you are seeing a performance drop both in 2D (on the monitor) and/or VR

2D performance seems fine for me - for example 70 fps on missions over Syria (btw I use the Steam version)

VR performance (running on SteamVR) is where I see the big drop, down to mid 20 fps for the same mission, lots of jitter and drops

Ryzen 9 5900X, 32GB DDR4-3200, with nVidia GeForce RTX 3080, HP G2 headset

[MattCri]

guys, it is not a windows issue if we can have the usual fps just reverting back the patch

 

in my opinion if you cant see the degradation you are possibly running 1440p with msaa or 4k with msaa on a very recent gpu, even less if 60hz or capped at 60fps

basically you dont see anything because your bottleneck is elsewhere

Edited November 4, 2021 by MattCri

[Enduro14]

Both 2d and vr on average was 10-15 FPS drop for me and in VR that hurts bad.  Plus can’t run same settings or Res compared to pre patch.  Don’t have numbers logged my test bed mission

 

Hind Syria free flight 

 

also to add I use quest 2 for vr, but still have rift s and use to run really nice settings in rift s.  I tested in rift s and cannot use same settings as pre patch without dropping below asw 40.  Also dropped the settings down to what I run in the quest 2 and still can’t maintain 40 in rift s even with those lower settings in current dcs version.  

Edited November 4, 2021 by Enduro14