February 20th, 2020 – Development/Insider Update

[Chock 17,765]

7 hours ago, sightseer said:

has anyone considered the GPU cycles per grass blade?  would you rather have great grass and a slide show or pretty good grass and good performance?  I don't really know how much of a performance hit each blade is but it can't be no hit at all.

There is a hit for multiple objects (or the appearance of them), but how much of a hit depends upon how the creation process for such displayed objects is done, thus how much work a GPU has to do to replicate displaying design.

In 3D - and indeed 2D - vector-based programs, there are essentially a few ways you can go about making something, but the gist of the difference between the smart way and the stupid way, is that you can either create a ton of individual 3D objects, which of course means loads of vector points need to be calculated (this is not a smart way to go about it), or you can create one object out of a few vector points and then basically state 'and when you've placed that one, display it again a bit to the left, and then again a bit to the left of that one, and then again, etc. Effectively 'rubber stamping it many times'.

A good example of this would be a model of a railway viaduct, where instead of modeling the whole bridge with each one of twenty arches having been modeled, you only model one arch and then have the program repeat the act of displaying it a further 19 times at specified distances from the original. Doing this obviously requires a lot less work to be calculated by your GPU since it has already done all the display thinking, and therefore only has to work out the positional difference data. The benefits of this are actually twofold; obviously it takes less time for the computer to think about drawing it, but it also makes the file sizes of the 3D objects themselves a lot smaller on a HD too.

Edited February 23, 2020 by Chock

[MatthewS 761]

1 hour ago, royalwin said:

The subtle runway imperfections are well below the noise floor of the DEM data

I'm not talking about "subtle runway imperfections", I'm talking about having the runway undulate based on the underlying mesh (and published data).

Edit: And you can't tell me that every dip/hump in runways are too small to be detected in the mesh... Sure maybe some dips/humps are below the "noise floor" but many will not be, I guess it depends on the vertical resolution of the elevation mesh.

Edit: A quick google search suggests 9m (24ft) vertical accuracy for SRTM DEMs. So maybe that is a little too coarse a resolution, but probably there are other datasets that are finer grained.  So maybe you're right...  But then, what data is XPlane using for it's undulating runways?

Edited February 23, 2020 by MatthewS

[bonchie 2,305]

1 hour ago, MatthewS said:

I'm not talking about "subtle runway imperfections", I'm talking about having the runway undulate based on the underlying mesh (and published data).

Edit: And you can't tell me that every dip/hump in runways are too small to be detected in the mesh... Sure maybe some dips/humps are below the "noise floor" but many will not be, I guess it depends on the vertical resolution of the elevation mesh.

Edit: A quick google search suggests 9m (24ft) vertical accuracy for SRTM DEMs. So maybe that is a little too coarse a resolution, but probably there are other datasets that are finer grained.  So maybe you're right...  But then, what data is XPlane using for it's undulating runways?

Xplane is using a very imperfect data set (probably just raw mesh) that produces really inconsistent results.

MSFS just isn't going to do that. If they can't get it right at all 37,000 airports, they aren't gonna do it.

Edited February 23, 2020 by bonchie

[Casualcas 830]

How many pages of this are there? I think after the first 5 comments, Asobo got it. 

What I'm glad to see are some questions answered.  In the miami shot, yes it looks like elevated roadways, not just 2d roadways,

I'm still (wetting) myself over the Donegal water

And more questions arise,  bloody h e  double hockey sticks, just how many different AI aircraft are there?

 The lens flare is gorgeous for screenshots, but when I'm flying my eyes dont do that, can I turn it off? 

Fuel loading is shown in gallons, can that be switched to weight? Pounds Kg?

Also hard to tell, if the airport striping is well travelled,  or freshly painted and pristine. 

Will the cargo area fit a nervous small Asian wife?

Edited February 23, 2020 by Casualcas

[MatthewS 761]

52 minutes ago, Casualcas said:

I think after the first 5 comments, Asobo got it.

You think?

Why then are we still seeing oversized trees in the latest video and yet that was raised a month or more ago ?!

[MatthewS 761]

3 hours ago, bonchie said:

Xplane is using a very imperfect data set (probably just raw mesh) that produces really inconsistent results

Ok, but if they are just using "raw mesh" then that means the vertical resolution is fine enough to exhibit some degree of undulation along the extent of the runway, otherwise it would just be a smooth slope (different elevations at thresholds only) or completely flat.

Edited February 23, 2020 by MatthewS

[tweekz 2,165]

3 hours ago, MatthewS said:

Ok, but if they are just using "raw mesh" then that means the vertical resolution is fine enough to exhibit some degree of undulation along the extent of the runway, otherwise it would just be a smooth slope (different elevations at thresholds only) or completely flat.

 

The kind of undulation we got in X-Plane actually make the experience less realistic. It's not rare that on my takeoff rolls (with jets) those undulations act more like a ski jumping simulator than a flight sim.

[ChaoticBeauty 6,692]

6 hours ago, MatthewS said:

Why then are we still seeing oversized trees in the latest video and yet that was raised a month or more ago ?!

On 2/21/2020 at 11:54 PM, ChaoticBeauty said:

It is very likely that the footage used in the episode is not from the most current build. This has been demonstrated with previous episodes, using footage from the early 2019 builds or even extended shots from the E3 and X019 trailers. It takes time to shoot, edit, render and encode an episode, so this is understandable.

The shooting and footage capture for this episode could have been completed several months ago. I mean, in the Global Preview event the first four episodes were previewed weeks/months before they were released to the public.

Or maybe they collected the feedback and have now put it on their extremely long to-do list, or they are working on the tree size as we're speaking. Things like these are not easy to fix world-wide, and they're not going to be fixed within one month from the feedback.

I wonder how you were feeling about Prepar3D progress back then. The transparent ATC window was promised in 1.3, and we didn't get it until 3.0. Yes, a transparent ATC window that Flight Simulator X already had. Microsoft Flight Simulator is progressing at light speed in comparison.

[MatthewS 761]

1 hour ago, tweekz said:

 

The kind of undulation we got in X-Plane actually make the experience less realistic. It's not rare that on my takeoff rolls (with jets) those undulations act more like a ski jumping simulator than a flight sim.

Hence the smoothing I mentioned in a previous post.

[MatthewS 761]

32 minutes ago, ChaoticBeauty said:

The shooting and footage capture for this episode could have been completed several months ago

Sure, but why show something so old? Have they not made any progress worth showing in those several months? Unlikely, IMO.

Edited February 23, 2020 by MatthewS

[suncoastflyer 685]

Only thing I can see majorly wrong here is lack of apron lighting. Other than that, this is going to be more than enough for me 95% of the time.

 

Gee just gotta say, in the original teaser trailer, LAX at dusk - the city looked very dark, no lights. Look at it now! That draw distance is jaw dropping O_O

Edited February 23, 2020 by suncoastflyer

[ChaoticBeauty 6,692]

33 minutes ago, MatthewS said:

Sure, but why show something so old? Have they not made any progress worth showing in those several months? Unlikely, IMO.

Because it takes time to capture footage and edit it. Work on an episode probably begins months in advance, and they'd rather use any existing footage than spend time capturing clips from a newer build, which would detract from development time. They are previewing development tools and specific details they're talking about, which means it is the actual developers assembling the footage for each episode, and they're not just outsourcing the work to a professional video editor. I am sure you will agree that it'd be better if they spent more time working on features we want rather than updating the footage more often. They are not demonstrating a finished product so it is not really necessary, and it's not like they even have to do any of this.

EDIT: And as the above post proves, there is evident progress when comparing older footage. Even if it lags behind a bit, our areas of criticism are being improved on.

Edited February 23, 2020 by ChaoticBeauty

[MatthewS 761]

15 minutes ago, suncoastflyer said:

Gee just gotta say, in the original teaser trailer, LAX at dusk - the city looked very dark, no lights. Look at it now! That draw distance is jaw dropping O_O

It does look good!

I hope a city's "glow" can be seen from a long distance (as in real life).

[Amon1973 1,401]

12 hours ago, MatthewS said:

Runways are long...

If you have a runway that's 2000m long and a 100m elevation mesh resolution then that's 20 elevation data points along the length of the runway.

Also "merge in" data points from chart data (eg runway thresholds) and give these higher priority than the elevation mesh.

The runway can be rendered at a high resolution in game (1cm?) with those 100+ data points being "smoothed" (think bezier curve).

The result is a smoothly rendered, undulating runway based on the underlying mesh and published data.

This approach can be applied to the entire airport surface (runways, taxiways, aprons etc)... I guess you need to exclude "spikes" in the elevation data from hangers/terminals etc

I think Asobo and Microsoft know what they're doing, so everything will be in hand.

[MatthewS 761]

8 minutes ago, ChaoticBeauty said:

I am sure you will agree that it'd be better if they spent more time working on features we want rather than updating the footage more often.

Don't they have over 120 employees? I'm sure one of them has time to capture footage from a recent build, rather than use something that is months old...

Edited February 23, 2020 by MatthewS