What We Want (Part 1)

[SteveW 1,792]

The way I see it is that all the multi-threaded mumbo-jumbo is kept well in-hand by the pros.

Looking into flight models, these generally have a certain granularity of CFD calculations, to give flavour to the flight model. But in terms of fast simulation, ideally these calculations are based from a set of lookup tables derived from more lengthy analysis.

Multi-threading is improving as hardware suits the code. Bearing in mind that as the core count increases, the use of 'all cores' for some tasks, may continue to be counter productive.

Most GPU functionality has a heritage in Microsoft software foundations and their research with GPU manufacturers and partners, arriving at DX10 preview in FSX. Later, these important times can be seen to be paving the way for programs beyond FSX, onto programs like XPlane that now have the luxury of porting to Vulcan.

[Paraffin 2,197]

5 hours ago, bashope said:

I think nowadays simulators' relatively poor performance comes from poorly optimized coding for multicore usage, inefficient coding practices and frameworks which tie up flight model calculations with the graphical part. I think that with experienced developers on "gaming" (or performance sensitive graphical applications, if you prefer) would be able to:

• Create a proper graphics engine with optimized "LOD" which can have great detail up close (like grass, moving trees, etc), but be able to quickly "switch it off" in the distance for performance, like pretty much any decent game nowadays.

I agree with the premise, but would just caution about comparisons with current games. Most games that use tricks with LOD to balance detail up close vs. detail in the distance don't operate under the same parameters as a flight sim, due to the speed of movement through the scenery.

If I'm flying a jet at low level through a major city like Seattle, which can be simulated in Google Earth's flight sim mode, there is a huge amount of 3D objects being loaded into the view. There is far more change happening in each animation frame than in a car racing game where the speed through the scenery is lower.

Only an air combat game like Ace Combat 7 comes close, and if you look carefully at the trailer for that game, there isn't a heck of a lot of detail being shown on the ground. Buildings are low poly count, everything uses low-res textures. 

I'm not saying optimized LOD isn't possible, just that the challenge is greater in a non-arcade flight sim if we're going for high detail and fast frame rates at 4k resolution. And one thing that will absolutely break immersion is if any of those LOD fade-in tricks are too obvious, like pop-in or texture resolution changes. All of that has to be hidden from view, pre-loaded in the background or hidden with distance haze to avoid breaking immersion.

[bashope 112]

31 minutes ago, Paraffin said:

I agree with the premise, but would just caution about comparisons with current games. Most games that use tricks with LOD to balance detail up close vs. detail in the distance don't operate under the same parameters as a flight sim, due to the speed of movement through the scenery.

I would refer again to the video I have shared before, from Unigine... it's possible, it's just a matter of skill (and not hardware), in my opinion.

[Murmur 3,943]

It is often heard the argument that flight sims show a larger area compared to other games and _therefore_ they can't have the same performance. But the CPU/GPU doesn't care at all if the rendered/visible area has a radius of 1 km or 100 km. What counts is only the actual number of triangles that must be calculated/rendered, the quantity/size of textures, and the lighting/visual effects applied. Other games might have less distance to show, but also a lot more detail up close.

Now it would be an interesting metric to compare the number of triangles and, secondarily, the total size of textures managed by a visually complex AAA title, vs a mainstream flight sim (say over NYC or London), and see the relative performance. Now that could give an idea of the relative efficiency between the two engines.

[Superdelphinus 718]

9 hours ago, Paraffin said:

I agree with the premise, but would just caution about comparisons with current games. Most games that use tricks with LOD to balance detail up close vs. detail in the distance don't operate under the same parameters as a flight sim, due to the speed of movement through the scenery.

If I'm flying a jet at low level through a major city like Seattle, which can be simulated in Google Earth's flight sim mode, there is a huge amount of 3D objects being loaded into the view. There is far more change happening in each animation frame than in a car racing game where the speed through the scenery is lower.

Only an air combat game like Ace Combat 7 comes close, and if you look carefully at the trailer for that game, there isn't a heck of a lot of detail being shown on the ground. Buildings are low poly count, everything uses low-res textures. 

I'm not saying optimized LOD isn't possible, just that the challenge is greater in a non-arcade flight sim if we're going for high detail and fast frame rates at 4k resolution. And one thing that will absolutely break immersion is if any of those LOD fade-in tricks are too obvious, like pop-in or texture resolution changes. All of that has to be hidden from view, pre-loaded in the background or hidden with distance haze to avoid breaking immersion.

Is that actually true though? It may be true that in real life there are more objects flying through view at a quicker rate, but in a game wouldn’t the amount of polygons being thrown at the screen be completely independent of the type of game? E.g. you could have a racing game with an extremely high polygon count around a deceptively simple oval track, and a flight sim with many fewer polygons making up more objects. The racing game would be the more demanding to process, despite it travelling slower and displaying ostensibly fewer objects. Im not sure if I’m explaining what I mean very well. I think basically my question is, is it actually relevant what type of game it is? What’s important is how complicated the objects displayed on the game are. Flight sims always seem to have very basic objects to me, even if there are a lot of them.  

[Superdelphinus 718]

56 minutes ago, Murmur said:

It is often heard the argument that flight sims show a larger area compared to other games and _therefore_ they can't have the same performance. But the CPU/GPU doesn't care at all if the rendered/visible area has a radius of 1 km or 100 km. What counts is only the actual number of triangles that must be calculated/rendered, the quantity/size of textures, and the lighting/visual effects applied. Other games might have less distance to show, but also a lot more detail up close.

Now it would be an interesting metric to compare the number of triangles and, secondarily, the total size of textures managed by a visually complex AAA title, vs a mainstream flight sim (say over NYC or London), and see the relative performance. Now that could give an idea of the relative efficiency between the two engines.

Oh, you’ve already said the same thing in a much more straight forward way!

[Noooch 1,084]

Would you mind staying on topic guys?

[bashope 112]

1 hour ago, Noooch said:

Would you mind staying on topic guys?

It seems on topic to me. We might be speculating and digressing a bit, but in the end we are saying we want a good engine and scalable performance in the new sim 😊

Personally I like the idea of adding to the flight model, making it more precise if enough resources are available. And of course, it must be very well optimized for modern CPU's, which are getting more and more cores, unlike the single-thread heavy simulators that we are used to (be it X-Plane or ESP based sims).

To me, as a programmer by profession, it seems even rather simple that the different aspects of the flight model (as shown in Austin Meyer's video) could be calculated independently in different cores/threads and added up in the end to calculate the new position and axis rotations... but hey I'm just speculating 🙂

[Wobbie 938]

Of course, comparing any MS core based sim to XP is like comparing Windows to Linux in a way, as, although both are sims, both use vastly different engines.

BTW ESP based sims are only applicable to P3D, the 'not for entertainment' sim, that stands on it's own.

Correctly you should have said... be it X-Plane or FSX based sims. 

[ifly777 4]

You know how in fsx the runways would "appear" or suddenly look more defined while on final approach? I would like it if in fs2020 they become defined more gradually

Edited September 10, 2019 by ifly777

[G-RFRY 1,712]

If flight simulators in the future set to higher benchmark for hardware you run the risk of loosing some of your market.

[aleex 551]

Volumetric lights illuminating the clouds. Something like PMDG immersion pack... 

Edited September 10, 2019 by aleex

[FlyerNYC 82]

I would really like good looking stock airports finally. The FSX/P3D ones are just garbage a mess. Would be nice not having to spend 25$ on every airport I want to fly to. I hope they are utilizing satellite images to construct them because it certainly looks that way if you manually brighten up the preview images of KLAX (at night).

Edited September 10, 2019 by SirGreenLemon

[carbonbasedlifeform 179]

On 8/29/2019 at 4:03 AM, bashope said:

Don't want to be rude, but I think your argument is invalid.

 

That is really impressive. I wonder if they could procedurally generate seasons. I assume Microsoft will have their own engine?

[DAD 2,653]

32 minutes ago, carbonbasedlifeform said:

That is really impressive. I wonder if they could procedurally generate seasons. I assume Microsoft will have their own engine?

Indeed, though the first 30 secs of the video pretty much reminded me to some of the MSFS2020 vids/shots seen from the MS team....