A new lease on performance

[OwenHewitt 0]

Since I am commenting on an almost 2 year old blog post, this may have been referenced long ago ... but since I have seen some recent comments regarding concern on performance, I thought that it was a worthy post. Please give it a read and give comments. Why FSX has Rough Performance.Regards,Owen

[Epgd 140]

Owen,thank`s for your thoughts.one question.... so why you didn`t say Sim? :)... Did you know something that we don`t..?brds

[Silverbird 115]

Since I am commenting on an almost 2 year old blog post, this may have been referenced long ago ... but since I have seen some recent comments regarding concern on performance, I thought that it was a worthy post. Please give it a read and give comments. Why FSX has Rough Performance.Regards,Owen

Thank's Owen it makes perfect sense and I do remember that post by Adrian wood's I'm glade some of the Aces team members were rehired for flight. since they have a good idea of what worked and what didn't work. I'm curios what type of file would be better then a mdl file? and what are the advantages of this other formant.

[Guest veeray]

Easy..... .X it's already being used by LM...

[OwenHewitt 0]

Owen,thank`s for your thoughts.one question.... so why you didn`t say Sim? :)... Did you know something that we don`t..?brds

There's always been some humor involved in calling this program for what it really is...a simulation. I guess MS is trying to really drive home that this is not just a sim anymore, so I thought that I'd have a bit of fun. I totally believe that this is a sim ... take a look at this other post I did.Best,Owen

[n4gix 4,947]

Easy..... .X it's already being used by LM...

It is? Funny I don't see anything but .mdl and .bgl files in MY Prepar3D installation......oddly enough, there's xtomdl.exe right there:

 Volume in drive Q is OS Volume Serial Number is 3CCB-9E57 Directory of Q:\Prepar3D SDK 1.0.2042.0\Environment Kit\Modeling SDK\bin10/16/2007  05:08 PM           527,712 modeldef.xml

[Epgd 140]

Owen, I was sure that answer..:) thank`sIt`s well known you have contact with some members of Flight team, (some of Aces).. maybe stupid question, but is there something, you know more about this project and can share with us?brds

[OwenHewitt 0]

Owen, I was sure that answer..:) thank`sIt`s well known you have contact with some members of Flight team, (some of Aces).. maybe stupid question, but is there something, you know more about this project and can share with us?brds

No sir, everyone is on the same playing field with this project ... I know just as much as everyone else or has been posted on the internet.It is a complete 180 compared to the development of FSX where team members were encouraged to blog about the project prior to release.Regards,Owen

[dmaher 0]

It’s a really interesting post, and it gives some good insight into the kind of problems faced.Right or wrong, I do get sort of exasperated when some people say Flight needs a new ’engine’; when to me that’s kind of wrongheaded.It’s the kind of surgery that could kill the patient. :( But I think ‘engine’ means a lot of things to different people.Torgo is talking about shaders (aka materials or shader objects+shader engine).They’re a small component but vital to performance and visual quality.FSX shaders are much more powerful than FS9.They’re also awkward, terribly finicky, and likely poor performers.But despite that, as far as I can see, FSX shaders are roughly equivalent to any reasonable DX9 game.They’re just not well refined for the artists.Rewriting the shaders to fix all that is no big issue.In fact, I’d only assume Flight will have new and improved shaders. (eng+obj)The problem is existing art. (planes, trees, buildings, grass - everything)All of this content needs to be re-exported to conform to changes, if the changes are significant.Worse, a substantial change may require an artist to reconfigure each shader in the model’s source file.That could be five minutes per model, or it could mean five hours…if new or altered texture maps are involved.The best time to make shader changes is early in the project.This doesn’t always happen...it can’t really happen that way.Changes need to be made if and when issues arise.Later in the project coders make every effort to avoid a mass re-export of the art.But, chances are, at some point the artists will get word they need to re-export global art.They just break into teams and get the job done ;)Actually in a well designed system this kind of re-export might be done automatically by script…if the changes are minor.And Torgo is saying, I think, the re-export problem could be avoided if the shader system was better designed to start with.FSX has a lot of legacy art - art that is carried over from previous games. It takes a lot of effort and attention to keep these old models working.And yes I’m sure coders took steps to avoid breaking this old content…and that may have limited the potential of FSX (...slightly).Late in the schedule FSX would need to live with their shaders good or bad.On the forums we talk about FSX’s old code…I actually think it’s this legacy art that's a problem.And not so much the engine ;)IMO the .mdl format isn’t going anywhere, it’s their format, and they can change it as required.There’s nothing special about it, just some names, attributes, and a ton of vertex locations.The shader changes mentioned do affect backward-compatibility for users…of course.But this shouldn’t be over emphasized.FSX aircraft and scenery would be re-exported with the newer SDK tools.There will likely be other changes needed – just like FS9 to FSX.But that upgrade was particularly difficult for a variety of other reasons.For the most part the source files for old can always be updated to work with new.(a big caveat is the scenery guys…there’s a lot of custom style data there)

[Silverbird 115]

It’s a really interesting post, and it gives some good insight into the kind of problems faced.Right or wrong, I do get sort of exasperated when some people say Flight needs a new ’engine’; when to me that’s kind of wrongheaded.It’s the kind of surgery that could kill the patient. :( But I think ‘engine’ means a lot of things to different people.Torgo is talking about shaders (aka materials).They’re a small component but vital to performance and visual quality.FSX shaders are much more powerful than FS9.They’re also awkward, terribly finicky, and likely poor performers.But despite that, as far as I can see, FSX shaders are roughly equivalent to any reasonable DX9 game.They’re just not well refined.Rewriting the shaders to fix all that is no big issue.In fact, I’d only assume Flight will have new and improved shaders. (shdr engine + shdr objects)The problem is existing art. (planes, trees, buildings, grass - everything)All of this content needs to be re-exported to conform to changes, if the changes are significant.Worse, a substantial change may require an artist to reconfigure each shader in the model’s source file.That could be five minutes per model, or it could mean five hours…if new or altered texture maps are involved.The best time to make shader changes is early in the project.This doesn’t always happen...it can’t really happen that way.Changes need to be made if and when issues arise.Later in the project coders make every effort to avoid a mass re-export.But at some point the artists will get told they need to re-export global art.They just break into teams and get the job done ;)Actually in a well designed system this kind of re-export might be done automatically by script…if the changes are minor.And Torgo is saying, I think, the re-export problem could be avoided if the shader system was better designed to start with.FSX has a lot of legacy art - art that is carried over from previous games. It takes a lot of effort and attention to keep these old models working.And yes I’m sure coders took steps to avoid breaking this old content…and that may have limited the potential of FSX.Late in the schedule FSX would need to live with their shaders good or bad.On the forums we talk about FSX’s old code…I actually think it’s this legacy art that's a problem.And not the engine ;)IMO the .mdl format isn’t going anywhere, it’s their format, and they can change it as required.There’s nothing special about it, just some names, attributes, and a ton of vertex locations.The shader changes mentioned do affect backward-compatibility for users…of course.But this shouldn’t be over emphasized.FSX aircraft and scenery would be re-exported with the newer SDK tools.There will likely be other changes needed – just like FS9 to FSX.But that upgrade was particularly difficult for a variety of other reasons.For the most part the source files for old can always be updated to work with new.

Very interesting Thank's for the info dmaher. :(

[Guest veeray]

And yes I’m sure coders took steps to avoid breaking this old content…and that may have limited the potential of FSX.

I think maybe that is the real complaint. You should never have a .MDL format where you have legacy support like that. Materials are obviously the big thing... the FSX .MDL supports compressed textures and decompression on the shaders.... why let anyone bypass this critical optimization?

[stonelance 136]

I think maybe that is the real complaint. You should never have a .MDL format where you have legacy support like that. Materials are obviously the big thing... the FSX .MDL supports compressed textures and decompression on the shaders.... why let anyone bypass this critical optimization?

MDLs don't do anything but reference textures. There is nothing built into MDL for texture compression. I'm guessing what you meant is that the engine shouldn't allow loading content that is not using compressed textures (including legacy content)? Texture compression always has down-sides to it (usually loss of quality), and you want to make sure that the artists (whose job it is to gauge the quality of the visuals) have the ability to decide when to use it and when not to. Obviously, this involves making sure the artists know what the trade off actually is (worse perf and memory usage), and usually the answer is 99.99% of the time you want to use it. That is one of the bad things done in FSX. No one really knew what was good to do and what was bad to do until it was too late to do anything about it.

[Guest veeray]

It'll be interesting to see what they decide on when it comes to Flight. In particular guages... I think the story will go similar to what we have with the texture materials. Perfectly good optimization with DX9 obviously, but do you lock out every aircraft that doesn't support it? Or do you wait until Flight II to do that?I think the latter.. although you can offer these optimizations now, they really don't become meaningful until the detail itself is increased.

[Christopher Low 5,359]

To be honest, I will be reluctant to purchase MS Flight unless it has full compatibility with the aircraft and scenery that I currently use. I am not too keen on the thought of spending the same amount of money on addons all over again. I am quite happy with the way the UK scenery and airports look at the moment on my PC. All I need in the future is superior performance.

[gmohr 0]

I specifically hope that existing planes/sceneries/etc are NOT supported. I believe that a huge reason for the poor performance and overal fragility of FSX is due to the baggage of backward compatibility. This is MS's golden opportunity to start fresh and create an efficient codebase with no sea anchors. I'll glady buy new addons if this is the case.