FS performance?

[hinch]

This is something I know many wish they knew...

Hoooow come games such as shoot 'em ups and racing games can have the fanciest graphics (all the water reflections and shadows you could ask for) but Flight Simulator seems to lag behind so badly? They all have physics, partical effects and much smarter AI.

I'm sure there's some good reasons but I can't justify it in my head.

Below I have a screenshot of Trackmania which came out last year with all kinds of special effects, 25 players on the screen at once etc etc and I still get 50fps with ease on my fairly modest setup. It's rendering fancy water, clouds, blurs, lovely light sheen on the tarmac, and loads of AI physics at the same time...

 

[toprob]

This has been covered in depth over at the Avsim forum. My own thoughts are this:
The screenshot looks like 1000 other games of this sort, making use of the 'basic' abilities of today's graphics systems to render water, and to use basic lighting such as bloom, specularity and bump maps. Which incidentally FSX uses.
The difference, I think, is when you look at how much is rendered there -- the hills in the distance are, what, a couple of kilometres away at most? What if it had to render the entire world, with a visibility of 100 kilometres -- from the air, rather than delineating the play area with handy hills? Somehow I suspect that your 50fps would no longer be possible. In fact, I suspect that with this type of engine 10fps would be impossible.

 

[arno]

I agree there, the two main differences for FS are that:

1. It has to render the entire world, the terrain that can be displayed is not a limited area
2. It has to be able to render it from any position, either flying at 30.000 feet or while taxiing on the ground

These two together are the main reason for what you call "bad" performance. I think FS is doing things quite smart though, also if I compare it to some of the image generators I use at work. There we might have more performance, but not with all the detail of FS (and the ease of going everywhere on the world).

 

[rightnumberone]

Arno,

Correct me if I am wrong, but FSX doesn't render the entire world at runtime. When I take off from KBOS, FSX is not spending computing cycles rendering the airport in Hong Kong, or 22,000 of the other ones. It is rendering KBOS and what surrounds me for a certain distance.

Having said that, I agree in general that there is no comparing these two games.

Just looking at the screen shot above, I'd say that the terrain mesh is certainly saving the dev some frames. Those clouds are .bmps that you can't approach and fly around in, and aren't moving (weather). The sky doesn't have to be corrected for time, season, etc. There is no passage of time. There are no other craft in the air (AI), in the water (boats) or on the highways. In fact, there are no highways.

The work that FSX has to do to render a realistic world is far and away more work than the CryEngine (or others of its type) has to do to create the little islands of world any given level presents to the user.

 

[Mace]

Arno,

Correct me if I am wrong, but FSX doesn't render the entire world at runtime. When I take off from KBOS, FSX is not spending computing cycles rendering the airport in Hong Kong, or 22,000 of the other ones. It is rendering KBOS and what surrounds me for a certain distance.

That's correct in essence (you're also getting region-wide AI traffic, region-wide season data, region-wide region data, etc.)

The problem is, the sim has to be ready to render KBOS from ground level, or from 10,000 feet. It has to be ready at a moments notice, to render KBOS from 2 miles out over Boston Harbor, to 5 miles out over the suburbs of Boston.

The sim doesn't have the luxury of most games to simply render one scene at one place in the world, or two or three places or views at once. It has to be ready to render from anywhere.

Too bad we have that situation...maybe some outside-the-box thinking can produce a better solution??

 

[rightnumberone]

Thought experiment

Mace,

Your post prompted me to try something. I've produce the attached image about 6,000 feet over Aukland, New Zealand. I set frames to Unlimited and conducted two experiments:

1) Autogen set to maximum
2) Autogen set to zero

I then measured (roughly) FPS and took a snapshot of the scene. I then spliced the two images together.

For an approximately 40% "price" in frames per second, I get part of this image.

Is the better part "worth" the 40% reduction in frames per second? Can you even detect where the image is spliced?

(Later on, assuming you can't detect where the image is spliced, I'll upload another image with the splice highlighted.)

Conclusion: Not sure I have one, except that maybe FSX charges a price for autogen generation at an altitude where the autogen does not signifantly add value to the scene.

 

[Timmo]

as well as the things said above, remember also that FSX has to try and render stuff from previous versions which bloats the code somewhat. It also has to render, in many cases, addon scenery and aircraft that may not have been made particularly efficiently.

 

[Mace]

Mace,

Your post prompted me to try something. I've produce the attached image about 6,000 feet over Aukland, New Zealand. I set frames to Unlimited and conducted two experiments:

Is the better part "worth" the 40% reduction in frames per second? Can you even detect where the image is spliced?

Neat thread!

I see the splice I would wonder about the same location at, say, 12,000 feet or 20,000 feet AGL; The autogen would probably not be there, or be so small as to not be rendered (SmallPartRejectRadius applying?)

Needless to say, I agree, there isn't much benefit to the user at those altitudes and above. But my thinking is that at some altitude FSX will stop rendering the autogen. I don't think 6,000 feet is the cut-off.

I mean, if you fly over Auckland at 30,000 feet, your frame rate will surely be better than if you're at 6,000 feet. At 30,000 feet the only major fps-killer will be AI in the area.

I am sure that ACES are thinking or will be thinking about these types of things for FS11 very closely.

 

[GHD]

I am sure that ACES are thinking or will be thinking about these types of things for FS11 very closely.

Hopefully they will remove the "popping" of autogen introduced in SP2

George

 

[bob5568]

the autogen fade in was removed to improve frame rates in sp1 if I remember right.

 

[GHD]

the autogen fade in was removed to improve frame rates in sp1 if I remember right.

You are probably correct Bob, however SP1 was a disaster for me, SP2 was the only saving grace. I still miss 'fade-in' of custom autogen when used with 1 m/p photographic scenery.

George

 

[Mace]

Hopefully they will remove the "popping" of autogen introduced in SP2

George

They (ACES) do know about alpha fade on the autogen, and our desire to see it in FS11. So I think they will do everything humanly possible to get it in there...at least as an option

 

[rightnumberone]

They (ACES) do know about alpha fade on the autogen, and our desire to see it in FS11. So I think they will do everything humanly possible to get it in there...at least as an option

Mace,

At about 11,400 feet, the autogen stops being rendered (I checked that too, while I was at it.)

I don't have the best rig in the world, so it could very well be that this was tested on top-of-the-line boxes and found to be an acceptable tradeoff.

My view is that the human eye can't detect autogen at 6,500 feet in sufficient detail to warrant the cost in frames. That's just an opinion, however ... and others may find it to be worth the cost.

 

[MichaelD]

I've had the exact opposite experience as George. This is more in line with the other thread going on but I had a very noticeable improvement with SP1. But SP2 has had some adverse affects, for me at least. I get the screen going blank for long periods of time or the graphics will suddenly change to huge pixelated squares (looks like FS4.0) and sometimes the computer will just reboot.

And this is on two *very* different computers. One running Vista with 4GB ram (although an older ATI card /256MB ram) and one running XP with 1GB ram and a GForce/Nvidia card with 256MB ram.

Amazingly, the lower end machine performs better with SP2 than the other but they still have problems that didn't exist before.

It's probably just me and I'm still tweaking things so it may get better.

-Michael.

 

[UtahJazzfan]

I also have great performance with SP1 but when I install SP2 real world buildings flicker. In Las Vegas anyway didn't try any other cities. I also lost about 2FPS with SP2.

Someone said autogen turned off at 11,400 feet. Is there a way to change this in fsx.cfg?