Some things to try for dual core users... and single co...

[Ramasurinen 0]

Hi guys. I've been doing a *lot* of tweaking since Sp1 came out (who hasn't?), and have been following the threads here, and I've come to the conclusion that the logical tweaking that one would expect to improve things doesn't really seem to, at least for me. I think I may have reached a point where I might be able to help some dual core users. Not sure, so maybe some of you can test this for me. Again, this is primarily for dual or quad core users but everything here, save my fibre frame setting, should help single core as well. I'll explain my logic as briefly as I can, which means this post will be long haha. ;) I have to state up front though that each of these changes I've made work in concert with each other. I encourage you to try them all, not just pick and choose, see what happens.First my specs. If you run something similar to this rig, who knows, what follows may help.Core 2 duo 6600 (moderately oc'ed to 2.99ghz)2 gigs 667 DDR2 ramnvidia 7900 GTO 512 pcie (not overclocked) on 94.24 whql driversA lot of us have been fighting with blurries since the release of sp1, myself included. Logic would state that increasing the fibre frame time fraction should help, but I've found that in the case of my core2duo it actually tends to hurt *far* more. Maybe Phil can correct me but I believe fibre_frame is a fixed value of time given to the scenery loader per cycle, whether the loader needs it or not, and is pretty much a direct drain on your fps regardless of whether that allotted time is used. Increasing this value can cause the rest of the rendering jobs to get delayed to the point that little else gets done on time. You might make a dent in the blurries, but what use is that if your autogen won't even show up above 200 kts? Once everything gets pushed too far down in the queue you're back to blurries anyway.Now for the first blasphemy. I have my fibre_frame set to 0.01. Yup, zero pretty much.As Phil mentioned, most of the threads for scenery loading have been moved onto the second core if you have one, so really what need is there for fibre_frame? It only seems to bog down core 0 in my case. This has given me a huge fps boost, but without the tweaks that follow, things can get a little nasty. Only dual and quad core users try this please. Single core guys, you need this value more than we do. Now to make up for any shortfall on scenery loader time caused by that, we need to reduce the strain in general on the renderer...Next up texture_multi_bandwidth. Logic says to increase it if you're getting blurries. Nope. From all the testing I've done I don't believe this helps terrain textures in the slightest. Back in the FS9 days (I love saying that) we had to increase this value to avoid the dreaded white/brown ground tiles, but ground textures appear to be done under another process now, at least on dual cores. What it does affect is how much in the way of textures on unique city buildings and aircraft that are coming into view are allowed to be loaded per frame. Basically texture_multi is a stutter factory. Keep it below 40, or...Blasphemy #2: I set my TEXTURE_MULTI_BANDWIDTH to zero. ;) The downside is that unique buildings (not autogen or terrain) sometimes appear briefly translucent and without texture when they come into view. Might drive some nuts, but doesn't bother me, and stutters over cities are gone. Basically with this setting I'm telling FS "don't stop for anything, I don't care how big and shiny it is."Another two things I've done that have helped enormously with performance, much more so than they did pre sp1. Use Rhumbaflappy's reduced default.xml file. I'm sorry I don't have a link to it, and not sure if it's still here on avsim. It didn't make much difference for me pre-sp1 (the bottlenecks were elsewhere), but wow did it ever after that. Thanks Rhumba! I think it was on Matt Fox's site awhile back, but maybe someone can provide a link to it here.Next, use one of the several reduced autogen tree texture sets available. I quite like TreeX 2.1. Use the low resolution set. Really a tree only needs to look perfect the second before you crash into it. ;) Even low and slow, the low res trees don't stand out, and give a nice performance boost. Fill rate still matters. They also don't shimmer with high anisotropy settings.Definitely set some limits on your autogen rendering in fsx.cfg. I have autogen density set to extreme but use these values in the cfg:TERRAIN_MAX_AUTOGEN_TREES_PER_CELL=800TERRAIN_MAX_AUTOGEN_BUILDINGS_PER_CELL=300This still gives me loads of autogen but also stops the great mountain forest fps hole from happening.And here's a BIG one. Back up your autogen folder somewhere safe, and then make two duplicates of AutogenDescriptions_Min.spb. This is the lowest variety set of autogen of the three options. After backing them up, remove AutogenDescriptions.spb and AutogenDescriptions_Mid.spb. Rename your duplicates of the MIN autogen file to those names. Now you have the same autogen set no matter what your sliders are set to, and it takes one hard target variable out of the fps equation. It works brilliantly in concert with rhumba's default.xml file.I've learned through all this tweaking that it isn't the amount of autogen on the screen that wrecks framerate but the amount of variety in buildings and trees. The less the game has to load in variety the more instancing it can do and the better the performance. As always, no crying if you don't back up these files before you mess with them. It's your sim and your responsibility. ;)On the topic of the dualcore hotfix in WinXP, I disabled mine through the throttle registry setting. I did nothing for me but cause microstutters. No performance difference with or without it.As for the once again popular mipbias settings, I tried it, but no difference for me except introducing shimmering. To kill the shimmering required turning on texture clamping in my video driver and that just made the videocard work harder for no visible gain. However, I strongly dislike photoscenery of any sort, so I don't use any and can't say if it helped that case or not.There is one BIG caveat to all these tweaks: FPS lock. The way I have this set up, if my fps drops below my framerate limiter more than a frame or two it can turn into a fairly nasty stutterfest. I've taken out the "padding" so to speak, so if something doesn't fit into each cycle it can get a little jittery. On the plus side, these tweaks have reduced my framerate volatility to around 1 fps. My FPS is extremely consistent for any given situation now. Essentially I'm saying set this up and tweak it using your worst case scenario and flying your most frame sucking aircraft. I will also add that I never use AI traffic, except for cars and ships, as I fly almost exclusively in multiplayer, so I have no idea what effect that will have. In conclusion... No single one of these tweaks was the proverbial silver bullet for my FSX, but I can state that I don't get persistent blurries anymore unless I am really abusing the sim, ie: 100 feet over a major city, with multiple friends flying in front of me at 600 knots in multiplayer, and with fraps running. I run all scenery sliders full right except for water, which I keep on low 2.x in multiplayer. Other aircraft on screen are too much of a hit with max 2.x water on. Flying solo I use max 2.x.The kind of flying I do is pretty brutal on the sim, 90% fast military at low level, and usually recording with fraps to make my crazy fsx films, so I think it should definitely help those who are less demanding on it.Again, don't bother with the fibre_frame stuff if you're single core, you need it. Dual/quad core users though, please give it a shot and please let me know here if you have any success.Cheers, and best of luck to you all. Let them eat cake...and have sharp textures. :)Mike

[Reset XPDR 1,271]

Mike,I thought I'd give your tweaks a try. The FPS certainly went up, as evidenced by a comparison run of SP1 stock using my FSX dynamic benchmark, BUT the side effects were quite noticable to me. http://forums.avsim.net/user_files/172977.jpgBlurries were particularly bad once the FPS dropped below 20 (around the 4 min mark) overflying Seattle and then again overflying KSEA. Also, whilst the FPS boosted considerably heading for Mount Rainier, the tree coverage was correspondingly considerably more sparse, which was not to my liking. On the positive side, the graphics were very smooth, with no stutters noticed at all :)And for those wondering what all this talk of SP1-related blurries is about, the following two pictures are before and after shots just over Boeing Field. The before picture was the image as it was around the 5 min 30 mark of my benchmark. I immediately paused the sim and waited 5 seconds or so for the blurries to go away and then took the second shot.http://forums.avsim.net/user_files/172978.jpghttp://forums.avsim.net/user_files/172979.jpgThanks anyway for taking the time to post your tweaks. At the very least is it interesting that our machine specs are so similar, yet blurries worsened for me with your tweaks, whereas they reduced for you. It must be the trademark FS blurries / stutter / FPS random generator at work ;)Gary

[Guest MauiHawk]

>Now for the first blasphemy. I have my fibre_frame set to>0.01. Yup, zero pretty much.>>As Phil mentioned, most of the threads for scenery loading>have been moved onto the second core if you have one, so>really what need is there for fibre_frame? Phil also reported that there *is* still tile texture work that needs to be done on your first core, regardless of how many cores you have... it may not be much, but its not nothing. Fiber_Frame, to my understanding, is only a guaranteed minimum time tile that tile work will get on the first core. Your locked fps will mean that tile work is still allocated on the first core anytime a frame is rendered faster than the time specified by your fps slider.I don't think that's necessarily a bad way of doing it... and when I install my dual-core chip tomorrow, I had actually planned on experimenting with low Fiber values (not *that* low, but maybe I'll play with that as well). Anyway, my guess is that the key here is that if you are going to zero out your fiber_frame value, you need to be sure your fps lock is actually limiting fps on a fair amount of frames to ensure the first core is getting some tile processing time. Otherwise you'll likely end up with a situation similar to what RESET is reporting.

[Guest jdeane]

Well - although SP1 seriously improved things, I was still experiencing fairly low frame rates (10-15)in urban areas, and when clouds were majorly present - but no longer !!! Your tweaks (fibre frame on 0.01, texture_multi_bandwidth on 0 and the two autogen reductions) have given me a steady 25 fps (locked) in every environment, with almost no blurries, AND with almost every slider on the right. So - it may not work for everyone, but it has certainly worked for me. Many, many thanks - and worth trying for anyone else with a dual core processor.Jeremy Deane

[Ramasurinen 0]

Awesome. Thanks for testing that out guys. Sorry it didn't work for you Reset. The performance graph was very interesting though.And yes you're right, you need to have some overhead available, several frames worth above the fps lock for this to work. As I mentioned in my post, if it just barely hits your lock or dips below things can get ugly fast haha.There's probably a sweet spot for fibre_frame per system, but I wanted to make that post just to show that FS can respond nicely to a little counterintuitive tweaking. It really depends on your flying style too I guess, but it seems that higher isn't always better for dual/quad cores. I'll keep on experimenting. 0.01 is probably a bit low indeed, and I've had good results with 0.05 and 0.1. edit: RESET, would you be willing to try again with your fps lock set to 25 fps? I noticed in those screens that you had it set to unlimited. I mentioned in the original post that for these tweaks to work you pretty much *have* to have your fps locked to a conservative number.Cheers.Mike

[PhilTaylor 45]

the fact that your machine "catches up" after some seconds would seem to indicate the settings are too high and its just falling behind.from your sig, it appears you have a lot of add-ons and your settings are pretty high. if you remove all add-ons, go to a minimal config ( single monitor, simple joystick, no TrackIR ) and choose the minimal perf bucket ( using the dropdowns ) does this problem persist? What if you disable audio?

[PhilTaylor 45]

Out of curiosity, what are your AA settings? Graphics card control panel set to "application control", and AA checked in the FSX dialog?

[Reset XPDR 1,271]

Hi Phil,The benchmark I run is with settings at detected default (global high across the board) on a single monitor, plus I turn on in-game AA and AF. I will do as you suggest and disable all addons and report back on my results. Before I do, when you say minimal perf bucket, do you mean the absolute lowest global default settings (eeewwww!)? In any case, if I still see blurries, I will cut back all throttles and see what she does!Gary

[Guest MauiHawk]

RESET,Read below-- Haldir's right. Your graph indicates you have your fps slider set too high (likely unlimited).For Haldir's tweak to work you must be sure the fps slider is capping fps to guarantee that texture work will be done. It won't make a very good fps chart comparison, but hopefully overall FSX will run smoother and clearer.

[Reset XPDR 1,271]

Ironically, I intentionally set the FPS slider to unlimited because of criticism I received for running a benchmark with an FPS cap set! In any case, when I was playing around trying to clear the blurries up when I first noticed them, I did a run with the limiter set to 30, then 20. As expected, the blurries kept at bay whenever the FPS counter was pegged at the limit set, but crept in as soon as it was challenged by the scenery. Even with it set to 20 FPS, I got blurries around KSEA with Haldir's settings.I will try again shortly, once I work out how to undo TileProxy that I tried for the first time today and has been giving me hours of entertainment (blurrie free too!). Phil, if you are still listening to this thread, get on to your Mircosoft bosses right now and hook up a satellite feed contractor that provides consistent data to pump scenery tiles straight into FS11 (or the FSX DX10 patch if you are feeling really ambitious). IMO, it's as revolutionary as real world weather was when it first came to FS! :-jumpy Gary

[Ramasurinen 0]

Haha, if future version of FS go photoscenery only, this will be the last version I ever use. ;) Suspension of disbelief goes out the window for me with that stuff.Anyway, RESET I find it kind strange looking at your specs that FSX is giving you so much trouble. In fact with your overclock and better ram your system should be slightly ahead of mine in performance. Maybe the extra couple mb of cache ram on the 6600 makes more difference than I thought, or it's a motherboard or video driver difference or something.One other tweak I've done that I forgot to mention was to reduce my prerender limit in the video driver when vsync is on to zero by using rivatuner. Have you tried that? I noticed when running at a prerender of 3 frames, the default driver setting, that FSX performed slightly worse in fps and blurries than when it is set to zero. It's a marginal difference for me but maybe worth a shot. You can get Rivatuner from www.guru3d.com if you need it. ATI users can make the same change using Ati Tray Tools, using its "flip queue size" function, again set to 0. This isn't functionality that is normally accessible through driver settings. It can also be done through the registry if you know where to look.Cheers,Mike

[Reset XPDR 1,271]

>Haha, if future version of FS go photoscenery only, this will>be the last version I ever use. ;) Suspension of disbelief>goes out the window for me with that stuff.Have you tried it? This is the first time my wife has looked at FS and actually knew exactly where I was flying because it looked so real! Anyway, I don't think it should be compulsory (especially with its insatiable download bandwidth appetite), rather a starring optional (deluxe version?) feature. >Anyway, RESET I find it kind strange looking at your specs>that FSX is giving you so much trouble. In fact with your>overclock and better ram your system should be slightly ahead>of mine in performance. Maybe the extra couple mb of cache ram>on the 6600 makes more difference than I thought, or it's a>motherboard or video driver difference or something.Actually, I think the help squad has jumped in somewhat unnecessarily early on this one, as FSX SP1 has been quite a positive experience for me (other than trying to work out what that second core is really working so hard at doing for seemingly little benefit). I only started seeing major blurries when I tried your settings / tweaks, and since going back to my settings it has been all goodness for me. Gary

[Ramasurinen 0]

Well thanks for trying them out anyway Gary. I didn't expect them to work for everyone, though it was cool to see it helped at least one person, and I hope others at least give it a shot.On the photoscenery thing, yes I've tried it. If photoscenery was 1m per pixel or better, and had full autogen placement, was properly matching and colour corrected and had all the natural shadows removed by hand, allowing FSX to shade mountainsides with its own dynamic shadowmaps, AND with proper dawn/dusk/night blending, then I might consider it. A nearly impossible task to be sure, but I'm demanding. As it stands today it just looks out of place to me. Painted aircraft against a photoreal background hurts my eyes. For me it's about consistency. If everything in the scene looks like it was made by the same hand then I am easily transported into that world regardless of its quality. If the pieces clash I just can't fall into it.I built and textured environments for various console games for ten years, so that's probably where my perspective on it came from. Yeah I know I'm weird. ;)Cheers.Mike

[Guest craig1998]

no your not weird...thats cool that you did that for counsel games.anywhoo I tried your stuff and Im generally impressed overall.the game runs pretty stable with no BIG stutters.Im wondering if the xml lite and autogen hax had anything to do with it as well since I did everything at once. but thanks for your effort and keep on tweaking is the name of this game.FSX---For stutters click X

[tf51d 284]

Well these are my results!Normal SettingsGraphics SettingsUnlimited FramesTrilinear (AF at 8x quality in graphic driver)AA off (4x with temporal and adaptive AA on in driver)Global Textire Res Very HighLight Bloom and Lens Flare Off, Advanced Animations on.Aircraft settings Maxed with 3D VCScenery settingsLevel of Detail Radius MaxedMesh Complex 100%Mesh Res 10m (FSGENESIS Mesh)Texture Res 1MWater Effects 2xlowLand Detail Textures checkedScenery Complex Extremely denseAutogen Density Dense (600 trees, 800 buildings per cell defined fsx.cfg default.xml not renamed)Ground Scenery Shadows offSpecial Effects detail HighWeather Maxed, (Use ASX now)Traffic 100% airline (Large modified file, using UT and WOAI flight plans)fsx.cfg settingsfibre frame set at 0.23Texture Bandwidth mult at 250trees per cell 600, buildings 800afinitymask is 3Result is smooth flight, good performance, no stutters, and no blurries.Test resultsSetting Bandwidth Mult to 0 results in good performance, no stutters, no blurries, but the momentary transparency of the textures, particularly when panning.settings fibre frame to 0.01 with unlimited frames, (Bandwidth back to 250) results in smooth flight, good performance, no stutters, but extensive blurries. Settings Fibre to 0.01 locking at 25FPS results in significant loss in frames from 20's at JFK down to 13-14FPS, no stutters (Other than loss of FPS) and extensive blurries.Settings Fibre to 0.23 locking to 25FPS results in significant loss in frames from 20's at JFK down to 13-14FPS, slight stutters (Not too bad though) and no blurries.All these tests flown in same aircraft (Overland A320) and weather conditions (Fair skies) same time of day and location (JFK 31L flying runway heading 2500 ft at 250kt's for about 15 minutes.