Frame issues in Prepar3D v4.5

[Darcy13]

I too was affected by 4.5. Short pauses (not stuttering) when taxing and turning and fps drop. Back to 4.4 and fps and short pauses gone.

thanks and look forward to the fix

Darcy

[Rasterfahnder]

@simon747 I like that attitude in your post. Could you maybe try to set the sim to unlimited and then compare 4.4 and 4.5? Because as you said it is possible to fly in v4.5 very well. It just takes a minute to uninstall the 4.5 client (dont deactivate your licens when asked) and install v4.4 for the test. As simbol pointed out the more the merrier.

I did now several flights from KFHR to KRNT (my personal test track since FSX, 14:00 lokal time, fair weather) with out framelock or vsync and for me the framedrop is real. So fare Ultimate Traffic Live, Orbx Global and Vector did play nicely with v4.5 so I keept them for my test. But even with just them running the frames with v4.4 startet at the runway of KFHR 78-108, in between arround 80 and on approach 56-84.
With v4.5 at the runway 62-85 in between arround 60 and on approach 27-48.

And I think thats maybe related to prepar3d. When you add ORBX LC, Regions and a decent weatherengine it gets worse of course.

[Alan_A]

EDIT: for those (like myself) who lock to 30Hz Vsync (and don't max all graphics settings) then you might not notice any issue

I do both of those things but I'm still seeing degraded performance - both fluctuations (rapid cycling from 30 to 15 to 30 fps) and the sim struggling to deliver 30 fps in circumstances and settings where 4.4 was able to do that consistently. Noted for what it's worth. Grateful for your work in running this down systematically.

@martyson - I ran the default scenario without Track IR and performance improved - I was even able to restore autogen vegetation to dense (one step up) and didn't see the stuttering I'd seen with Track IR running. But I honestly think those results are an artifact, not a cause. Panning around in Track IR puts a heavy demand on rendering - much moreso than when you're using the joystick hat switch (and as a result you're panning a lot less). My results with changing autogen settings tells me that something's off with rendering in the core sim - and Rob's results indicate that that's not add-on dependent. So am hoping his results get... um... results.

[Martyson]

EDIT: for those (like myself) who lock to 30Hz Vsync (and don't max all graphics settings) then you might not notice any issue

I do both of those things but I'm still seeing degraded performance - both fluctuations (rapid cycling from 30 to 15 to 30 fps) and the sim struggling to deliver 30 fps in circumstances and settings where 4.4 was able to do that consistently. Noted for what it's worth. Grateful for your work in running this down systematically.

@martyson - I ran the default scenario without Track IR and performance improved - I was even able to restore autogen vegetation to dense (one step up) and didn't see the stuttering I'd seen with Track IR running. But I honestly think those results are an artifact, not a cause. Panning around in Track IR puts a heavy demand on rendering - much moreso than when you're using the joystick hat switch (and as a result you're panning a lot less). My results with changing autogen settings tells me that something's off with rendering in the core sim - and Rob's results indicate that that's not add-on dependent. So am hoping his results get... um... results.

"I ran the default scenario without Track IR and performance improved - I was even able to restore autogen vegetation to dense (one step up) and didn't see the stuttering I'd seen with Track IR running. "

Thai is good news.


I no longer use TrackIr (previously mentioned).
I have two monitors (std 60Hz).
An old PC.
No maxed out sliders.
Window mode with Ezdok.
ActiveSky (after its Beta).
Orbx (west) regions and a few Orbx, FSDT, FlightBeam airports.
A few payware aircraft.

I lock FPS at 20.
FPS stays at 20FPS. No jumping around.
For aircraft that "demand" PMDG/F1 B200 e.g. and using add-on scenery I shut down e.g. Copilot PFD.
All ok for my needs FSX/sp2, LM/P3D (all versions).

[Alan_A]

I lock FPS at 20.

Per Rob's post, with that setting and those additional restrictions, you're probably fine in 4.5. At a 30 fps lock, I'm just hitting limits, but 20 sound like enough overhead. Enjoy it! I'll stand by for a fix. Different styles and preferences - nothing wrong with that at all.

[Rob Ainscough]

Thanks Rob for the test , in a way it kind of proves its not an ORBIX problem but a problem with the new LM 4.5 update as i dont have any kind of Orbix scenery and i see the FPS problem ..

Agree, it does seem to be an issue LM need to investigate and resolve as I don't think a 29% drop in FPS from V4.4 to V4.5 (with NO add-on products installed) was an expected result. LM dev team is very responsive to key issues like performance so I'm sure they'll make this a priority and provide a solution.

Rob.

[CaptKornDog]

LM dev team is very responsive to key issues like performance so I'm sure they'll make this a priority and provide a solution.

Glad to hear that.

[Martyson]

Notes:
For all my normal flights on my older system (FSX / P3D) , I use Locked 20 FPS.
I do not have Orbx Germany South / North so cannot do test there.
I also posted this test on the Orbx forum:
https://orbxsystems.com/forum/topic/171 ... th/page/2/

My quick test, P3dv4.5 (Client, Content, Scenery) 2 monitors

KSFO with Orbx and Flightbeam , PMDG 737NGX, SODE Jetways in use at all gates.


Locked 20 VC cockpit view, non gate parking = 19 to 20
Locked 20 External view, non gate parking = 19 to 20

Locked 30 VC cockpit view, non gate parking = 29 to 30
Locked 30 External view, non gate parking = 29 to 30

Unilmited VC cockpit view, non gate parking = 45 to 49
Unilmited external view , non gate parking = 58 to 61



Locked 20 VC cockpit view, gate parking, SODE Jetways, 19 to 20
Locked 20 external view, gate parking, SODE Jetways, 19 to 20

Locked 30 external view, gate parking, SODE Jetways, 29 to 30
Locked 30 VC cockpit view, gate parking, SODE Jetways, 29 - 30

Unilmited VC cockpit view, gate parking, SODE Jetways, 39-42
Unilmited external view, gate parking, SODE Jetways, 55 to 67

[AnkH]

Agree, it does seem to be an issue LM need to investigate and resolve as I don't think a 29% drop in FPS from V4.4 to V4.5 (with NO add-on products installed) was an expected result.

Rob, did you port over your shadow tweak from v4.4? I realized that using "4096" for those two parameters:

OPAQUE_SHADOW_TEXTURE_SIZE=
TRANSLUCENT_SHADOW_TEXTURE_SIZE=

which gave me wonderful crisp shadows in v4.4 (VC and close surrounding of the plane) totally ruins performance in v4.5. With cloud shadows active, my FPS droped from locked 30 to about 10-12FPS (!!!) for example above KDEN from Flightbeam and cloudy weather. Unticking cloud shadows -> bam, 30FPS again. Using "1024" instead of "4096" for both parameters above -> 30FPS even with cloud shadows ticked.

Might be that LM changed the shadow map quality or how it is rendered? Might be that due to the increased need of VRAM some observed in v4.5, the above tweak is simply to much for my 1080Ti? Who knows...

[Kayla Dillon]

Hello all,

Thanks for all the feedback. We're currently trying to hunt down what caused this. Any further information is always appreciated and helpful.

Regards,
Kayla

[Rob Ainscough]

Rob, did you port over your shadow tweak from v4.4? I realized that using "4096" for those two parameters:

OPAQUE_SHADOW_TEXTURE_SIZE=
TRANSLUCENT_SHADOW_TEXTURE_SIZE=

My process for testing:
1. Verify all 6 P3D folders don't exist, delete them if they do
2. Install V4.4
3. Manually set graphic settings via P3D UI
4. Exit P3D
5. Run P3D
6. Set location and aircraft and time of day 12:00:00, save flight file
7. Record test flight (.FSR) and save file
8. Save prepar3d.cfg for reference only
9. Exit P3D
10. Load MSI AB and enable logging to file (OSD is optional)
11. Run P3D
12. Load Flight file
13. Load .FSR saved recorded flight
14. From an external PC I record the flight ... this is a 4K capture card that records direct from HDMI splitter so as to NOT influence "load" on the main FS test PC
15. Exit P3D after recorded flight completes
16. Save MSI AB .HML file with logging data (CPU usage, GPU usage, FPS, frametime, etc.) (really a CSV file)
17. Save video capture of flight on external PC to file (MP4)
18. Copy .FSR recorded flight file to Documents (root) so it will not get cleared out for V4.5 testing
----------------------------------------------------------------------------
19. Uninstall V4.4
20. Delete 6 P3D folders
21. Install of V4.5
22. Manually set graphics settings via P3D UI (check they match same settings as V4.4)
23. Exit P3D
24. Copy the .FSR recorded flight file from Documents (root) to P3D Files folder
25. Load MSI AB and enable logging to file (OSD is optional)
26. Run P3D
27. Set location and aircraft and time of day 12:00:00, save flight file
28. Load .FSR saved recorded flight
29. From an external PC I record the flight ... this is a 4K capture card that records direct from HDMI splitter so as to NOT influence "load" on the main FS test PC
30. Exit P3D after recorded flight completes
31. Save MSI AB .HML file with logging data (CPU usage, GPU usage, FPS, frametime, etc., this is CSV format)
32. Save video capture of flight on external PC to file (MP4)
----------------------------------------------------------------------------
33. Bring the .HML (CSV) files into MS Excel
34. Filter out beginning and ending data that is not part of the actual recorded flight (401-403 frames of actual data)
35. Use Excel to calc AVG, Max of FPS, Memory, etc. in order to provide a summary
36. Calculate Delta (percent change) between V4.4 and V4.5 using this formula; (((y2- y1))/ y1) * 100

Rob.

[simon747]

@simon747 I like that attitude in your post. Could you maybe try to set the sim to unlimited and then compare 4.4 and 4.5? Because as you said it is possible to fly in v4.5 very well. It just takes a minute to uninstall the 4.5 client (dont deactivate your licens when asked) and install v4.4 for the test. As simbol pointed out the more the merrier.

I did now several flights from KFHR to KRNT (my personal test track since FSX, 14:00 lokal time, fair weather) with out framelock or vsync and for me the framedrop is real. So fare Ultimate Traffic Live, Orbx Global and Vector did play nicely with v4.5 so I keept them for my test. But even with just them running the frames with v4.4 startet at the runway of KFHR 78-108, in between arround 80 and on approach 56-84.
With v4.5 at the runway 62-85 in between arround 60 and on approach 27-48.

And I think thats maybe related to prepar3d. When you add ORBX LC, Regions and a decent weatherengine it gets worse of course.

Hi I already tested in the same test spot and the fps are the same in 4.4 vs 4.5 in my test spot for this time at EGLC. No difference in Orbx global and uk2000 eglc with only dense autogentrees, no buildings. Very smooth for me on med settings.

[keenrw]

@simon747 I like that attitude in your post. Could you maybe try to set the sim to unlimited and then compare 4.4 and 4.5? Because as you said it is possible to fly in v4.5 very well. It just takes a minute to uninstall the 4.5 client (dont deactivate your licens when asked) and install v4.4 for the test. As simbol pointed out the more the merrier.

I did now several flights from KFHR to KRNT (my personal test track since FSX, 14:00 lokal time, fair weather) with out framelock or vsync and for me the framedrop is real. So fare Ultimate Traffic Live, Orbx Global and Vector did play nicely with v4.5 so I keept them for my test. But even with just them running the frames with v4.4 startet at the runway of KFHR 78-108, in between arround 80 and on approach 56-84.
With v4.5 at the runway 62-85 in between arround 60 and on approach 27-48.

And I think thats maybe related to prepar3d. When you add ORBX LC, Regions and a decent weatherengine it gets worse of course.

Hi I already tested in the same test spot and the fps are the same in 4.4 vs 4.5 in my test spot for this time at EGLC. No difference in Orbx global and uk2000 eglc with only dense autogentrees, no buildings. Very smooth for me on med settings.

Trouble is you may not see the FPS issues running at Dence trees and no buidings etc so your frames will be more or less the same , i run max buildings and trees but medium draw distance and the the FPS drop from 4.4 to 4.5 is noticable which kind of makes me think its the added autogen buildings into 4.5 that may be causing this.

[simon747]

Yes, sorry, I rarely use autogen buildings, but use autogen draw distance one from highest setting and and dense speedtrees. Not getting a drop, looks like LM are looking into for those ppl who are having a drop..

[rfelger88]

I too am having FPS issues since updating. I have noticed that it improves greatly after disabling any weather program. I was using Active Sky with 4.4 & had no issues. Since the update the FPS dropped by15- 20 or so with weather loaded. All my settings are the same as in4.4