Dynamic Lighting Performance & 4.1

[vgbaron 3,685]

38 minutes ago, OzWhitey said:

Now, I'm not a tech guy, but this doesn't make sense to me based on what I'm hearing people report.

You, like everyone else, are entitled to your opinion. I find a tad laughable though that a self proclaimed "not a tech" guy would disparage findings by those who ARE tech guys. Since you seem to rely on what you read, should you care to, I suggest you research NickN and his relationship to ACES and the ESP engine.

As to the "must do" - I agree that was to broad - there have been changes in P3D, off loading things to the GPU, that make it not as important. However, anyone who is having problems related to slow loading textures which also affect FPS have nothing to lose by trying it. Even the most reliable of tweaks do not work for everyone.

Vic

[OzWhitey 3,353]

11 minutes ago, vgbaron said:

I find a tad laughable though that a self proclaimed "not a tech" guy would disparage findings by those who ARE tech guys. Since you seem to rely on what you read, should you care to, I suggest you research NickN and his relationship to ACES and the ESP engine.

I'm glad you find it amusing, Vic. I don't work in the IT field, but I do know a bit about science, so I tend to be interested in observations and data, as per my comments in this thread. Do I "rely on what I read"? Well, I think you'll find most science-y types get a lot of their data from reading, but they also assess the quality of the evidence presented.

Having been an avid simmer for several decades, I - of course - know who is NickN is. Like a lot of other people, back in the day I wasn't a great fan of the document that he - and apparently you - somewhat dubiously call 'the bible'. To be blunt, it's a pretty idiosyncratic piece of writing! 

For anyone who hasn't read this famous piece of flight simulation history, here it is, read it and make up your own mind: http://www.simforums.com/forums/the-fsx-computer-system-the-bible-by-nickn_topic46211.html.

As per LM's comments (from Robert McCarthy): "While Prepar3D was built off of the MSFS engine we have completely overhauled the engine from the ground up since we started development. From rewriting the rendering engine to support DirectX 11 to upgrading the platform to 64 bit and Unicode, the entire engine has been completely reworked and modernized."

So I wouldn't suggest that anyone assume that a technique - even if it did work in FSX - is a 'must' for P3D. Try out it, and see if it actually improves performance on your system. And if it's a situation like this - where there seems to be a problem requiring LM's attention - report the data to the community, there's a lot of issues that can be solved when large groups of motivated people work together. 

[AnkH 2,271]

Well, it does not make sense to argue about NickN and his bible. I tried that before, never made sense. Fact is, people really treat it as a bible and as it is with people taking the "real" bible as granted, discussions are normally useless.

BUT: I always find observations such as the panning thing interesting. Basically, you did exactly the same by switching on and off DL while flying and checking the FPS counter. It was an observation. You did not see any difference in FPS, but then? I observed that switching on and off DL inside the running sim does not really turn on and off DL, it leads to something strange in-between. In my case: if I fire up EHAM from FlyTampa with DL on, my FPS are terrible. Then I switch DL off, FPS are back to normal. Then I switch DL back on and FPS remain, but what I SEE on the screen is obviously different from what I have seen with DL initally on.

Means: observations are extremely individual. However, if several people observe the same (as they did with the panning stuff), it gets interesting. Personally, I do not really care about the scientific explanation behind, if it helps to restore the FPS back to normal when using DL, I will take it.

[gboz 157]

5 hours ago, vgbaron said:

As to why LM hasn't done anything to fix it - I'd speculate it's on a LONG list of fixes to be done down the road. Considering there is an easy workaround, I'd doubt if it were deemed a "critical" move.

Except there a number of threads over at LM with plenty of posters complaining about DL performance and LM has never mentioned this workaround in any of their replies. You would have to conclude that they do not know about it.

gb.

[Dirk98 354]

6 hours ago, airservices said:

This is one thing that I always did in FSX on the start

it was one thing in the NickN bible and I have used it ever since

Yes, same here, but I stopped doing that years ago when VAS became an issue.

[Guest]

3 hours ago, OzWhitey said:

Aren't simmers getting low frame rates sitting stationary at the airport? The necessary textures for the render for that position would be loaded into the L1 cache after a certain - very brief - period of time.

Your logic is sound however it's not that simple, and my supposition was just that (I don't do bibles and I encourage people to question everything and anything).  I don't have access to the source code and I don't have the necessary performance profilers specific to a GPU (nVidia or AMD) so I have no solid foundation of "fact" to demonstrate my supposition.  I only have many years (I've lost count) of coding experience from games to enterprise level business apps, I've been in and out of DX and OpenGL but I certainly would NOT consider myself a guru in 3D graphics programming, leave that to the LM development team.

If you look in some of the .DDS files for textures, you'll notice many different types of textures are contained in one file that can be applied to many "objects" or just "one object".  In addition there maybe diffuse maps, specular maps, bump maps, emissive maps, etc. that also need to be read/loaded ... and then there are Texture.cfg files defining fallback textures.  

So it's possible the low frame rate at a static view point is a result on not loading enough complete texture files to utilize the texture cache so the shaders are operating slower and having to do more round trips.  How the optimization is handled by the driver/hardware I honestly don't know.  This is just my theory, without code and nVidia or AMD provide profilers, it's just my speculation and nothing more.

There are also other factors that will determine how much of an impact DL will have on performance, some of that is down to how the scenery developer implements there DL lights.

When I was working the V4.x Betas I had made several anomaly videos for LM, I didn't understand the load and FPS numbers I was getting ... here are some of those videos:

In this 2nd video below I was trying to determine why I was getting an FPS drop  at 0:36 mark?

In all cases this was with 4XSSAA at 4K res with DL enabled.  I presented this to LM and they indicated the cause for the FPS drop is how lights are heavily coupled to scene complexity.  How far the DL light casts will have a huge impact on performance (airport beacons and landing lights is a good example for far casting), in addition, the larger lights that are closer to the camera/viewpoint the more per pixel work needs to be performed (similar to how particle effects will impact FPS, jet wash for example, when you stand directly behind the Jet wash and the screen is filled with particles, FPS plummets (especially with higher AA levels).

The discussion I had with LM did bear fruit, they are aware of the issue and they DO HAVE plans for future enhancements.  So we wait for those plans to come to life in a future version and/or update ... no ETA.

Cheers, Rob. 

 

[OzWhitey 3,353]

Just now, Rob Ainscough said:

(I don't do bibles

...says the man with 6,666 posts. 

[Guest]

Too funny, but I don't the devil either, I'm indifferent ... what's good today is bad tomorrow and what's bad tomorrow is good today ... change is the only constant I know.

Cheers, Rob.

[OzWhitey 3,353]

Whilst thinking about your comments in the previous post, I did a spot more testing.

I've been using the QOTS II up until now. This is on a clean install of 4.1, with a 1080Ti and a 5820K @4.4. 

Thought I'd try the A2A Commanche, as it's got a dynamic light in the cabin (must say I didn't notice this until recently!).

At KGKT Gatlinburg-Pigeon Forge (a mid-sized airport), I'm getting 78 frames per second with dynamic lighting off, with GPU at 46% load whilst sitting at the airport next to a hangar.

With dynamic lighting on in settings, and the landing lights and cabin light on, frame rate is unchanged, but GPU usage goes up to 51%. The results are similar whether I'm sitting on the runway or pointing my landing lights directly at a hangar.

There's a bit of subjectivity with the exact numbers reported, but effectively there's very, very little performance impact. This is radically different from those who've had frames drop into the single figures once the lights get turned on. 

Now, an early poster said that lack of performance degradation was because I wasn't using payware airports. That's part of my hypothesis. I may be having good fortune with the lights because I'm very picky with where I fly. Because I'm chasing performance - my typical use is flying with an Oculus CV1 in complex aircraft, like The Queen - I don't use payware airports currently, because I'm not happy with the performance I can get in this situation.

However, I also don't ever use default airports - yuk! Not sure I could bring myself to do that. :) 

Rather, I design my own airports with performance in mind - plenty of objects, but I'm increasingly careful to only use items from the P3D library. I've found that - dynamic lighting aside - the use of some older scenery onjects, that worked well in FSX/previous versions of P3D, now how a negative performance impact. 

So I suspect that some of the problems come from ported airports that have lots of FSX components in them. As I mentioned above, Mir alluded to something like this quite some time back.

In terms of what's behind all this, I'd be interested to know if anyone is getting severe performance impact from dynamic lighting when:

- they're at a default airport

- they're using a decent GPU (with VRAM that meets LM's specifications) 

- nVIDIA inspector is not being used. Edit: Also, not using SSAA in the sim.

- Shadows are off. 

Edit: Just went over to the LM Prepar3D forum. In the main thread there about dynamic lighting problems, I note that the OP found no frame drop once he left his specific payware airport (Samos), and tried the lights at a default airport. It was also noted that he had a video card below specifications.

One other suggestion I saw - that hasn't been mentioned here I think - is to turn off the 'high resolution textures' option. That's also how I currently have my sim set up.  

 

 

 

 

 

 

 

[simbol 5,821]

1 hour ago, Rob Ainscough said:

I presented this to LM and they indicated the cause for the FPS drop is how lights are heavily coupled to scene complexity.  How far the DL light casts will have a huge impact on performance (airport beacons and landing lights is a good example for far casting), in addition, the larger lights that are closer to the camera/viewpoint the more per pixel work needs to be performed (similar to how particle effects will impact FPS, jet wash for example, when you stand directly behind the Jet wash and the screen is filled with particles, FPS plummets (especially with higher AA levels).

The discussion I had with LM did bear fruit, they are aware of the issue and they DO HAVE plans for future enhancements.  So we wait for those plans to come to life in a future version and/or update ... no ETA.

Cheers, Rob. 

 

Hi Rob,

This makes perfect sense, I studied the DL lights this weekend for the PMDG 737 NGX (Which people complain a lot about bad performance with DL), users keeps indicating that the FPS drops heavily after turning the Landing Lights and the Taxi lights, the configuration of these effects shows a high quantity and distance of Dynamic Lighting.

My understanding from what I have read from some forums is that PMDG is blaming LM for the FPS drops (I don't have any links from PMDG sources), interesting enough some users implement the "pan around" solution and the performance comes back at normal levels.

The question is: Should PMDG adapt their DL lights to reduce the impact on the system and make their customers happier? or should LM try to address this issue?

Best Regards,
Simbol

[branks737 29]

1 hour ago, simbol said:

The question is: Should PMDG adapt their DL lights to reduce the impact on the system and make their customers happier? or should LM try to address this issue?

Best Regards,
Simbol

Can they at least meet half way?????? 

[Cargostorm 151]

28 minutes ago, branks737 said:

Can they at least meet half way?????? 

The half way already is that legacy lights work with the PMDG 600/700 expansion models 😗

[earthdog 842]

Just now, Cargostorm said:

The half way already is that legacy lights work with the PMDG 600/700 expansion models 😗

But with DL off i cant see lights on the ground textures in 737-700...Shouldn't i be able to see them?

[simbol 5,821]

33 minutes ago, branks737 said:

Can they at least meet half way?????? 

It is up to them.. what can I do? 

Simbol

[simbol 5,821]

2 minutes ago, earthdog said:

But with DL off i cant see lights on the ground textures in 737-700...Shouldn't i be able to see them?

The 737-700 / 800 aircraft model do not have night textures attached on the ground for this goal, instead it was designed to use Dynamic Lights, so if you turn DL off they are not there.

Simbol.