Is reality getting in the way of performance....

[PhantomTweak]

Maybe not, but the passengers near them sure can. One of my favorite passtimes on a long flight with very flexible wings, is to sit next to a nervous passenger, and then look out the window, gasp, and close the shade real fast. They look out the window, see the wings flexing, the engines bouncing around, the spoilers assisting the ailerons, and go into so much panic..

Lotsa fun to then lean back and go to sleep. Makes them even more worried. What if I'm just praying with my eyes closed? What if I just don't WANT to see the crash....etc :D :D

Works great until I snore. Kinda a givaway. Oh well

Pat☺

[truenorth71]

To try and answer the OP's question, I think first the question needs to be reversed: "Is performance getting in the way of reality?"

 

As developers are constantly trying to push the boundaries with regards to aircraft complexities, weather generation, and AI, we as the consumer have to try and keep pace with them, and I speak for a lot of folks out there when I say that if that's the case, keep the credit card fully stocked.

 

BUT....

 

Yes, everyone has a big BUT, as I see it, it depends on how far you are willing to go. I believe that currently, the flight simulation market is saturated with a good cross-section of quality software to create a very good reproduction of reality, as long as the consumer has the computing power to process it all. I have had people ask me what I use for hardware and software to achieve what I get in a sim and I immediately counter it with "what exactly do you want your sim to do?"

 

Sure, you can make pretty pictures with the sim, hang it on your wall and marvel at the achievement, only to realize that you couldn't get the plane to fly without your computer crashing. You can also flight an aircraft half way around the world in a single shot and to that end, sit back and marvel at the accomplishment all the while recognizing that you had minimal autogen running and no weather.

 

Getting back to my counter question, if people say, "I want it to do everything", well break out the wallet because it'll cost. I'm sure a lot of other folks here, myself included, have quite the collection going, probably adding up well over a grand. It's a hobby and with any hobby, it takes passion, dedication and of course, the sense of enjoyment.

 

Look at how FSX was coded, back in the day. It was way ahead of it's time and even nowadays, people are still having trouble, no matter what they throw at it. Since then, we have the Steam Edition version of FSX, P3D and X-Plane...the big three. I won't get into arguments about which is better, because in the end, we make a choice based on what we want. Some may go for all three and fly all three because they serve a certain aspect of flight simulation that the others don't. Some may only choose two of the three and then there are those who choose one and stick with it. Regardless of our choices, we can agree that we made them based on what we want out of a sim. The field will expand a little more now with DTG's products, AeroFly and whomever else is developing, and so we continue to review, research, discuss and choose. Debates will continue to ensue...Coke/Pepsi, Apple/Android, Mac/PC, Ginger/Maryann and all the while, the companies still make a buck off of us because they know that every time they come out with the next iteration, version or even a new product, someone will get it, hoping they get one step closer to reality.

[Erick_Cantu]

The big problem, as Hiroshi Igami pointed out to me, is that most developers are more concerned with detail than accuracy or performance. This isn't a way to say that detail is bad, but rather, that most developers aren't concerned with rendering things efficiently (and some take pride in massive polygon counts). Let's take one of my favourite examples - I really like the Captain Sim 737 model. The inlets are way too small, and the vane flaps are curiously missing (perhaps the extra effort went into the flight attendant's chest area), but overall it's the most accurate model of an early 737 (when compared to my engineering data from Boeing) presently available. The external model alone is 200,000 polygons. It could have been done in 60,000 and nobody would have ever known the difference. The way the windows are done is a perfect example - not only are the windows themselves modelled, but so are the lips around them. This could have been done with a texture and nobody would have ever known.

 

I've been told that PMDG used alpha channels to do the cabin windows on their NGX. I looked at some high-res screenshots, and you really can't tell the difference, so kudos to them for staying in the realm of reality. This comes with its own caveats - the only way to do this that I know of is to tie the reflection intensity to the specular alpha, so the texture artist does lose some control (why we don't have an opacity map is beyond my comprehension, but this is the same team that gave us a thrust reverser animation that only really makes sense for the levers, so whatever). There is always going to be a trade off, just like real-world aviation.

 

To answer your original question, triangles that are not presently being rendered aren't really going to slow down the rendering of ones that are being rendered. That your framerates drop when a complex terminal or other structure comes into view, for example, if you're in spot view rotating the camera around the aircraft, and go back up when the background returns to empty space is a fairly basic indicator of this. There may be some minor performance hit associated with loading a model file that is physically larger.

 

The problem with textures is a nebulous composite of resolution, number, and UVW coordinates. While larger images do take up more memory, it is always more efficient to have a single large texture than many small ones. A balance must be struck. The ideal MSFS aircraft in this day and age should be composed of one to three 4096 x 4096 pixel bitmaps (as opposed to a hundred smaller ones). You could get very good resolution on just about any commercial jet with three big images. I was feeling nostalgic last night and decided to download a bunch of repaints of the v2 FlightFX DC-9s (SGA in name only, after all!), and the texture layout was absolutely horrific to me (and how come nobody pointed out that the -10's gear bays were positioned wrong?). If you compare it to the later 727, you will find that the resolution of most parts is comparable (in some cases greater), yet the number of images is far fewer. These days one ought to stuff as much into the empty spaces between parts on a smaller number of massive images as one can get away with (while keeping the painter's job from becoming too onerous - after all, the painter's work is what makes or breaks a model). The problem then becomes one of reducing the number of UVW coordinates, and most model builders are far too lazy to be concerned with all of the welding and whatnot that goes into that (often myself included). I think modelconverterX has some automated ways of doing this (as well as generating LODs!). To me, it seems as basic as running MSTS models through shape fixer that one ought to give every final release model a run through modelconverterX to minimize drawcells and create LODs.

 

Honestly... I don't think that too many developers really put any real thought into either efficient rendering or texture layouts (the situation has greatly improved with respect to basic dimensional accuracy, but the problem of model builders filling models with imagineered details to make things more impressive to the average simmer who doesn't know an aileron from his arse persists!).