A first glimpse at Prepar3d v2.3

[KingGhidorah 209]

"You're overclocking...if you have to overclock to get proper performance, your pc does not have plenty of horsepower"

 

Ed, I don't even know where to begin with that one.  One's "horsepower" is about what somebody ends up being able to crunch with it, not about what it's base specs say, on paper.  And since the person you were responding to has a 4930 to begin with, and a Titan card, I would say that he does indeed have plenty of horsepower, especially with an overclock.  I consider not overclocking to be a waste of value of your processor, and if you've seen software problems caused as a result of an overclock than either there is something wrong with the software, or something wrong with the overclock. 

[Guest]

Gents, I am sorry to come down on this side of the fence but it is my opinion that that the stutters that are apparent in P3D V2.2 are solely as a result of Vsync and FPS. I have successfully run P3D V2.2 with the setting low enough to maintain a frame rate 60fps on a 60hz Plasma TV and flown around for over 20 mins with Zero stutters and perfectly smooth even in at the horizon in a step turn. The odd thing for me is that will experimenting to achieve this I locked the FPS to 60 and gradually reduced the setting until I achieve a constant 60FPS. Then I changed the FPS to Unlimited and voila the Frame rate is over 170 FPS. Nearly 3 times higher at the setting that are required to get 60 Locked. Now there has to be unrealized potential there? Right?

 

Anyway at these setting with the monitor at 60hz there are some intermittent stutters locked at 60FPS with Vsync off. There are lots of stutters at around 170 FPS, But there are no stutters at 60fps with Vsync enabled.

 

I made a video with ShadowPlay that proves it and I will post a link to download the MP4 file. The mp4 File does exhibit the odd stutter that was not apparent will I was actually making it but I still think the video has sufficiently long 100% stutter free perfectly smooth footage to make the case. I won't put it on YouTube because that process is guaranteed to add lots of stutters:-)

[Guest]

Here ya go dave, I did this over the weekend in 2.3 (you'll need to watch this a 1440p to see the usage values):

 

 

You can download the MP4 from my Google drive if you want: https://drive.google.com/uc?id=0Bw0Q-fAfEZwyVzRLSl8xa0tTS1U&export=download

 

The upper right shows the following:

GPU Temp © | GPU Usage | Frame buffer load | Local Bus Usage | GPU Core Clock

GPU Mem Clock | GPU Mem used

CPU 1 to 12 usage

RAM (this is combined VRAM and VAS)

FPS (also in the upper left)

 

The exact Display settings and nVidia settings are here:

Upper Left: http://robainscough.com/P3D_2_6.php

Upper Right: http://robainscough.com/P3D_2_7.php

Lower Left: http://robainscough.com/P3D_2_3.php

Lower Right: http://robainscough.com/P3D_2.php

 

I think what you and other's might be missing ... look very closely at the stutters.  Suppose I'm running around 30 fps and my monitor is 60 Hz ... so I should be missing 1 frame ever 1/30th of second.  This wouldn't be a big problem if I was missing exactly 1 frame every 1/30th of second (for many not noticeable and my feel "fluid").  The problem is that is NOT happening ... they are NOT evenly missed frames.

 

What IS happening is more like this (over the duration of 1 second):

1/60th = 0 frame render

2/60th = 1 frame render

3/60th = 0

4/60th = 0

5/60th = 0

6/60th = 1

7/60th = 1

8/60th = 0

9/60th = 1

10/60th = 1

11/60th = 0

12/60th = 1

13/60th = 0

14/60th = 1

15/60th = 0

16/60th = 1

17/60th = 0

18/60th = 1

19/60th = 0

20/60th = 0

21/60th = 0

22/60th = 0

23/60th = 1

24/60th = 0

25/60th = 1

26/60th = 0

27/60th = 0

28/60th = 0

29/60th = 0

30/60th = 0

31/60th = 1

32/60th = 1

33/60th = 1

34/60th = 0

35/60th = 1

36/60th = 0

37/60th = 1

38/60th = 1

39/60th = 1

40/60th = 1

41/60th = 0

42/60th = 0

43/60th = 0

44/60th = 1

45/60th = 1

46/60th = 1

47/60th = 0

48/60th = 1

49/60th = 0

50/60th = 1

51/60th = 0

52/60th = 1

53/60th = 1

54/60th = 0

55/60th = 0

56/60th = 1

57/60th = 1

58/60th = 1

59/60th = 0

60/60th = 1

 

The above represents 30 fps (30 frames render in one second).  However they are NOT render in a uniformed manner so the eye picks them up as "obvious" stutters.  Now if that pattern were exactly 1 frame render then no frame render then 1 frame render then no frame renders, etc. etc. in that exact order for the entire duration of 1 second then the eye will most likely not interpret a "stutter".

 

Version 2.0 of P3D seemed to be much more uniform at rendering frames over one second.  

 

A stutter is a longer duration between when one frame is render vs the next frame (in above example 25/60th to 31/60th) ... that longer delay in rendering frame triggers the eye/brain to see a stutter.

 

I believe this is the problem LM are facing and trying to address in 2.3.

 

Cheers, Rob. 

[alulekk 3]

Awesome detail on the analysis Rob!!

[WarpD 824]

Say what???

 

Not sure I agree with that conclusion at all. In fact, when watching the video... I see two distinct types of stuttering. One caused by a total pause in the render loop and one caused by rendering being out of sequence with the video display.

 

While the one that interrupts the actual render loop can be eliminated software wise... the other one can not.

[gboz 157]

Thanks Rob, that's a lot more scientific.

 

It's pretty easy to test the 60hz vsync theory.

Just lower your settings to get well above 60 and engage it.

 

Doing this in my case I can "prove" that the stutters still remain.

 

As Rob indicates the stutters in P3d are far greater than anything

that would usually be produced by not using vsync.

 

gb.

[Guest]

Doing this in my case I can "prove" that the stutters still remain.

Yes on your system. But doing this on someone else's system

where the stutters are entirely removed demonstrates that it

is a system issues and not a program issue.

[gboz 157]

Yes on your system. But doing this on someone else's systemwhere the stutters are entirely removed demonstrates that itis a system issues and not a program issue.

Yes it's starting to look that way.

But meanwhile just telling people to turn on vsync is not going to work for a lot of systems.

 

gb.

[Guest]

 

 

I see two distinct types of stuttering

 

Actually I see 3 types of "stutters".

 

 

 

But doing this on someone else's system
where the stutters are entirely removed demonstrates that it
is a system issues and not a program issue.

 

Not true at all, it just demonstrates a synchronization issue where one system just happened to hit the timing correctly while another did not and since it's rare to find two identically performing systems with identical graphics settings and identical add-ons in identical locations ... that's why we see such variance in "stutters".  The issue is about threading and that's why AffinityMask changes make a difference to some and not to others ... 2.3 defaults to ALL cores.

 

If it weren't a coding issue, why would LM invest time in trying to reduce stutters?

 

Cheers, Rob.

 

EDIT: I need to get P3D V2.0 loaded as I think that will help demonstrate it's not system related.

[bonchie 2,305]

 

This Oxbx post says

 

 

 

and

 

Isn't that all about cloud shadows?

 

He said the cloud shadows now have a smaller impact on FPS, so he did comment on performance aside from just checking it out. That's what I was pointing out.

[Guest]

Did some additional performance testing with a comparison with P3D V2.0 using similar graphics settings.  V2.3 Beta appears to have about a 10-15% performance drop in terms of fps.  Again full clean installs no add-ons.

 

NOTE: After uninstalling V2.3 beta, the V2.0 installer failed as it appeared to think V2.2 was still installed ... apparently the uninstall for V2.2/2.3 leaves some key registry entries that the V2.0 installer finds ... it was easy enough to resolve by looking at the V2.0 installer log ... there are some GUID references that I had to manually remove from the registry before the V2.0 installer would work.

 

Here is a video of V2.3 at KSEA:

 

Graphics settings are here: http://www.robainscough.com/P3D_2_5.php

 

Here is a video of V2.0 at KSEA:

 

Graphics settings are here: http://www.robainscough.com/P3D_2.php

 

Again, keep in mind this is a 2.3 "early" beta.

 

Cheers, Rob.

[mike2060 4]

Rob, you can always use Revo Uninstaller to completely remove P3D. It will first use the program's uninstaller then will scan for folders and registry entries associated with P3D.

 

http://www.revouninstaller.com/revo_uninstaller_free_download.html

[corkyo4 2]

Gents, I am sorry to come down on this side of the fence but it is my opinion that that the stutters that are apparent in P3D V2.2 are solely as a result of Vsync and FPS. I have successfully run P3D V2.2 with the setting low enough to maintain a frame rate 60fps on a 60hz Plasma TV and flown around for over 20 mins with Zero stutters and perfectly smooth even in at the horizon in a step turn. The odd thing for me is that will experimenting to achieve this I locked the FPS to 60 and gradually reduced the setting until I achieve a constant 60FPS. Then I changed the FPS to Unlimited and voila the Frame rate is over 170 FPS. Nearly 3 times higher at the setting that are required to get 60 Locked. Now there has to be unrealized potential there? Right?

 

Anyway at these setting with the monitor at 60hz there are some intermittent stutters locked at 60FPS with Vsync off. There are lots of stutters at around 170 FPS, But there are no stutters at 60fps with Vsync enabled.

 

I made a video with ShadowPlay that proves it and I will post a link to download the MP4 file. The mp4 File does exhibit the odd stutter that was not apparent will I was actually making it but I still think the video has sufficiently long 100% stutter free perfectly smooth footage to make the case. I won't put it on YouTube because that process is guaranteed to add lots of stutters:-)

+1 to this.

 

I have an i3 processor @3.30Ghz. I had awful stutters and attributed this to the weedy processor (even with minimal settings in p3d). However since applying adaptive vsync @ 30fps with a monitor refresh rate of 60hz the sim has been nothing but silky smooth when the fps are at 30. Whenever I drop to around 20fps I get stutters, now at 20FPS I doubt the human eye would tell the difference in image smoothness between 20-30fps, i.e. the stutters that area apparent around 20FPS are due to vsync. so I am convinced the issue is the synchronisation between images being produced from the GPU and the refreshing of the monitor.

 

Tom

[Matthijs 79]

 

Did some additional performance testing with a comparison with P3D V2.0 using similar graphics settings.  V2.3 Beta appears to have about a 10-15% performance drop in terms of fps.  Again full clean installs no add-ons.

That's very interesting, many thanks for testing. Hopefully they are able to fix it in a new beta version. B)

Edited July 1, 2014 by n4gix
Removed excessive quote!

[guenseli 1,591]

 

 

 V2.3 Beta appears to have about a 10-15% performance drop in terms of fps.

 

 

Sorry, english is not my native language: "performance drop" means performance is reduced 10-15%?

Or is there a gain in perfomance?

 

 

And another question: I have currently (v2.2) only bigger problems over large cities (namely city autogen). Do you, Rob, see any improvements especially there?

 

many thanks