Hi,The title may sound like an infomercial, but this is an essay to analyze the issues surrounding FSX performance for many of us, and an attempt, through analysis of the technical factors involved in the rendering pipeline, to find a solution.Many reports I

"How does this translates in FSX now? After making a lot ofresearch on the retail version, I came to the conclusion thatFSX, as compared to FS2004, is HEAVILY relying on video cardvertex and pixel shaders to render what we see. In fact,EVERYTHING passes through a shader of some sort and most ofthe time, a single pixel on the screen passes through between2 and 3 different pixel shaders at least. Add on top of thatvertex shaders to deform/transform geometry, and you alreadyhit the common 4 shaders pipeline common place to most ofvideo cards today. DX10 standards, in my opinion, aresupposed to make a minimum of shades mandatory for the videocard, and I suspect FSX is designed with this

You might want to read some of the other threads in the hardware forum before you conclude. There may be some value in what you say - lessening the imapct on the video card - but you can always make a horse run more slowly by cutting off its legs. The challenge is to keep as much of what makes FSX FSX and not just `FS9 in a different box`, and make it work in acceptable ways - in other words, to whip the staggering nag back into a galloping steeplechaser. What you propose basically emasculates FSX to no more than `FS9 + addon` levels - and we have already got that, thanks. So why would anyone do that?Lets deal with the first factual error in your statement: Contrary to what you suggest, VISTA is NOT, DX10 only. DX10 is Vista only, but that is a very different thing. Vista ships with a new variation of DirectX9, and that is what we are told FSX was developed for. So we cannot let you make assertions regarding performance under DX10 without asking you prove it, which of course you cannot do as no-one here has seen FSX (or any game in the retail market for that matter) running under Vista and DX10. So please keep those `conclusions` which are only speculations to yourself, until you can prove them. And you have no benchmarks about Vista/DX9 and FSX, so really you're trying to run before you can walk. But good experimenting so far.What we'd really like to see is detail from the ACES team of precisely what `optimisations` for Vista/Dx9 FSX has included? Together with comparison benchmarks on identical hardware with XP/Dx9. And not on beta or release candidate versions of either software, with lashed-up drivers and poor excuses, the Final Version of Vista and the release version of FSX. For that, we need to wait a bit longer. So just continue with FS9 in the meantime, leave FSX well enough alone - and solve the problem!Your next factual error is claiming that DX10 cards perform faster than DX9 cards. How can you know this since not a single card has been run on public demo, with benchmarks? It is my understanding, based on contacts through work, that there were serious performance issues with DX10 cards, when asked to run DX9 applications - which is what FSX is, after all. There is no market acceptability for a high performance card that costs a high price while actually delivering reduced performance on non-DX10 games, which is why there are no firm delivery dates or performance indicators for the next gen cards yet, which you would have expected by now if the information was favourable. In fact, ATI have yet to actually announce a DX10 card, and NVidia have not confirmed that the G80 is DX10-optimised - these are conclusions drawn by the media, without foundation at this time. All very cagey.I know everyone would like to see a magic wand waved over FSX and suddenly 50 frames per second pop out, but it simply isnt going to happen until BOTH the hardware and the software have been upgraded to cope. It's a brave move by MS, but they need to hold onto the customer base while it finally accepts that todays hardware - and even Win XP - are simply not up to the job of getting FSX to do what it is supposed to. In twelve months time we won't be having these kind of conversations, but in the meantime I suggest that while you keep your experiments going, you don't publish the results as `conclusions` which have no basis in fact.Allcott

My understanding of DX10 from what I've read stated that DX-10 cards are required to follow a "unified" programmable shaderapproach (something ATI has already implemented and NVidia is due to release in their next cards in Nov.) Rather than having a fixed number of Pixel shaders and a fixed number of vertex shaders in each pipeline, DX-10's unified shaders will allow the DX10 programmer to use the "pool" of shaders any way they want/need to. So if you need more pixel shading but not as much vertex shading, you just configure your shaders that way.Regardless... I think doing an analysis like this is a good thing. However, we really need to wait until a wider range of people have their hands on the full product and can we can find a way to make a Database that captures both PC setups, driver versions, and FSX settings so we can determine what's going on here. Maybe it has a lot to do with which brand video card or CPU or how much memory you have, etc. Right now the reports are just WAY too varied and way too incomplete to really gain any good information. We need FACTUAL data, not just "impressions".--2002cbr600f4i

Absolutely right. There's already signs that this confusion is getting simmers confused.All we know is that we don't know. Yet.Given that Vista itself is gon to be an expensive upgrade, and the DX10 upgrade after that will definitely require new hardware, that's a certainty that whatever it takes to getting FSX working right is, it is going to be expensive! Allcott

Hi Allcott,I think this is very simple in fact: video cards in the market in 6 months, that will be DX10, will be faster than current video cards, as much as current video cards are faster than the ones released 6 months ago.What I write above is that DX10 cards, which inherently will be new cards from current gen, will then be as inherently faster than current gen DX9 cards. In short: DX10 cards will be faster than DX9 cards ONLY (Mainly?) because they are of a new generation.Speculation? I've took the time to re-code the bloom shader in FSX and I surely got a less visually attractive effect, but an impact of only 2 FPS instead of 12. This showed me reliance on shader.Furthermore, each polygon texture in FSX now can have Ambient+Fresnel+Specular+Bump+Light maps. That is 5 per pixel. Most hardware today handle at most 4 shader pipelines.I've not tried yet but a friend of mine is having more FPS with the F1 ATR than the default C172. My reasoning is that FS9 materials only use 2 textures per pixel per pass, while FSX is 5. Would it be 4 to fit most hardware today, maybe FSX would gain more speed.Nevertheless, I can't sum up days of experiment in a short essay in the forum, and I'm not trying to gather people under my sole vision. I've just spent time experimenting in putting the best of my knowledge to track the bottlenecks and find issues, and it worked very fine for me.An additional note: with the settings above, FSX is not emasculated and offers a richer experience than FS9 on my test system. Your mileage may vary.If only it helps more people having a better experience, I would then feel pleased not having wasted my time!

I would very much like to thank you for going to the effort of sharing your findings with us. I would also like to thank the respondees for their comments too.The value of this forum is that we can all share our various experiences, trial suggestions, and hopefully all benefit from what we offer one another.Downunder we have to wait until about 26 October for the release of FSX, but I am gathering information from those of you lucky enough to be experiencing FSX, and sharing it with us. I thought the transition from FS2002 to FS2004 was really good, and I think the potential of FSX is another leap forward.Thanks again.CheersMike

Certainly wasn't a waste of time for me, Jean! It helped a lot, and still looks great. Thanks a million!

Went from low teens to stuck on 25 my preset!Actually raised my autogen sliders without a hit.Still only a few flights so I do not know what all is missing or what the above actually did but right now Im happy.Keep up the experimenting!

Thanks JeanLuc. I trust that you have done ample testing before making such statements. I have followed you advice on many issues and you have always been on the mark. I trust you expert knowledge in this case as well.I purchased my Deluxe copy here in San Antonio yesterday (thanks to whomever gave the tip to get it at BB on 410 and San Pedro) and have found that FPS is acceptable and fluid in my scenery area (KSAT) if I make several adjustments to Autogen and Fiber_Frame_Time_Fraction. I'm still experimenting and plan to move on to more graphic intense airports soon. My system is not on the cutting edge: ATI Radeon X1300 video card with 256mb; 3GHz CPU; 1Gig RAM. I have a dual boot system with WinXP-SP2 and Vista RC1/5600. I have FSX on both XP and Vista with no problems in my scenery area.I really like the fact that we can now easily save configurations and load them when needed. I plan to create a JeanLuc config with the settings you suggested and test it on both WinXP and Vista.My next step is to use the ASv6 Wx Engine with FSUIPC 4/beta. I'm having a lot of fun so far with this new sim. I can't wait until the add-on market ramps up to full speed.Thanks a bunch JeanLuc. Keep us informed on any other findings you come up with. I am more than eager to try them out.Take care,

Thanks a lot Ken! You and the other now make me feel I don't loose my time LOL!I think I recall having seen the x1300 being twice as slower than the x1400 overall.I'm posting an addendum with some more findings in the thread now

Hi,further spending some of my sunday with this, I've made a couple more observations.First, about the settings I use, I've been able to push up to 18m for the mesh, but I find no real visual improvement for the 2-3 FPS less it gives. I'll stick with 76m for now and this suffice enough to me.Triliner/no AA: I used to set these in the Catalyst Control panel. However, it appears AA in Catalyst and AA in FS = blurry 2D panel. If I also set Anisotropic filtering in FS, I get shimmering textures. Reviewing some of the shader code in FSX, it appears some of the shaders are using point sampling, and I strongly think from the blurry in the 2D panel that dissapear when I uncheck AA in catalyst, that in fact, FS AA and Filtering settings are coded by shaders and supercedes any other settings in your driver control panel. A bit of an issue because I like using 2x AA for trading off performance/visual and maybe the FSX AA is using only 4x, hence lower FPS.Last but not least, you may also consider deleting the following folder that holds precompiled versions of the shader code:C:Documents and Settings--windows user name---Local SettingsApplication DataMicrosoftFSXShadersAnytime FS runs, it dynamically compiles the shader code and store it here ready to be sent to the video card. Deleting this folder will make sure any remnent of the now disabled "misc" folder will not interfere. So you delete it once, and from now on FSX will recompile the shaders in it and this will match your now disabled misc folder (no need to delete this Shaders folder again).That's all for now!

Further testing with the "optimized" autogen files found in this thread:http://forums.avsim.net/dcboard.php?az=sho...id=354861&page=It does give an additional performance increase on my test system as well. Better trying this one too as well!thanks Matt for these new files!Hope this helps!

Well, I keep trying!disabling entirely the misc shader folder introduces some artifacts with the way clouds refreshes when changing weather, and loose the environment mapping (for example on the engine spinners).you may instead consider renaming the misc.disabled folder back to misc, and instead, inside the misc folder, disable a couple shaders to compare:Bloom.fx -> Bloom.fx.disabledVolumeShadow.fx -> VolumeShadow.fx.disabledZOcclusionTest.fx -> ZOcclusionTest.fx.disabledthis is all experiment of course.Hope this helps!

Jean-Luc,Thanks for sharing your ideas/tests here. I really appreciate those who're intent on finding solutions and/or 'workarounds' to some of the performance issues while we wait for hardware to catch up with the demands FSX makes on our systems......it's a very refreshing counterpoint to those who's only "gift" to the forum(s) is incessant complaining... ;)However, I do have one request. PLEASE don't use that silly checkbox at the Message format line to use "plain text!" It makes it very, very hard to read that tiny blue text... ;)