I closed my eyes for a second tonight and saw something... like the 'flux capacitor'. No, I'm not going crazy... I really have put two and two together and come up with a relatively good theory on why FSX seems to bring higher end systems down to the performance of more archaic machines.Everybody knows that heavy duty processors from the last few years have all relied on having large CPU caches. This is why, as a gamer, I chose the 1meg cache version of the AMD 64 4000+ rather than the much cheaper 512k version. In terms of sheer processing speed, the larger cache enabled much higher benchmarks than the cheaper alternatives. However, the cache wasn't the only thing that was different.... the actual architecture of the processors were slightly different so that the 1meg version can fully utilize its larger cache. To say the least, if somehow that 1meg cache were damaged, leaving only 512k actually working, its performance would not be equal to, but much less than the full 512k version. I hope to god this is making sense so far.Anyways. Here goes the theory. Rendering the cache useless on any processor brings the raw processor speed down. If you take an older 3500+ processor with a 512k cache, and my 4000+ with it's 1meg cache, it doesn't take a rocket scientist to figure out who would post the best benchmarks. However, disable the effectiveness of the cache on those processors. The 3500 really wasn't gaining a whole lot of performance boost from its measly 512k, so it will still plod on with relatively decent throughput. The 4000+ would simply run into a corner and whimper, it raw speed requires a full meg of cache to really do anything at a decent speed. It's performance would drop and people would argue, but it is said that the larger the cache, the greater the hit. The 4000+ would probably benchmark the same, or even less than the 3500 even though the 3500 is no longer running at its full performance rating either.So how does this relate to FSX? Well... Doc saw the 'Flux Capacitor' and everything became clear.... I saw it too. FSX is using Fibers to run several tasks seemingly simultaneously between frames. When Win32 Fibers switch in and out, the CPU cache all but clears. Fibers renders CPU cache's useless. FSX uses fibers. FSX seems to run sweet on some (not all) lower end processors. Could this be why?From personal experience, I built a machine for a co-worker while I was still shopping around for parts for my own system. The result was that we both have the exact same machine, except while I paid a little extra for a higher performing 1meg cache, he was more than happy with 512k. We have ran FSX side by side, exact same drivers, exact same settings. His FSX vastly outperforms what I get.I suggest an experiment. If we could get simmers to respond here and post the following specs, maybe we could figure out if I should go back to the future and never return :)_________________________________________Processor : AMD 64 4000+ @2.6GHzL1 Cache : 64k + 64kL2 Cache : 1024kFPS : 15 - locked and maintained 80% of the timeFiber setting : 0.98You can find out your cache sizes using CPU-Z. Also, post what FPS your getting with whatever settings you deem as being flyable, and more importantly, whether it is locked or not. State whether it maintains that locked FPS and judge how often it does that.I think we will see a trend. Faster processors with larger CPU caches will probably perform similar to a lower end processor with a smaller cache. Just an experiment really... could be worse, I could be asking you all to take your temperatures and PH levels :)p.s. I am posting this for another reason. I want another smart IT person to come on here and tell me I've got it all wrong. I want that to happen because if indeed, fibers are destroying the usefulness of large CPU caches, then it would mean probably months before a patch would be released because it wont be an easy thing to fix :(Kev CPro IT ConsultantAmateur PhotographerPro Airbus Stress Tester/Impact Tester (kidding)

Good topic, and frankly I believe you are right. Another reason I still haven't bought into this marketed dual core technology crap.Anyways, CPU-Z is indeed a great tool to measure this, and to entertain you, here is my information.* AMD 3500+ Venice Core (Bus speed 220MHz x 11 multiplier = 2.42 GHz.) Yes, she is overclocked. Cache is 512Kb, naturally.* 1 GIG Xerox PC-3200 DDR RAM running at 220MHz with standard timings of 2.5,3,3,7.* ATI Radeon 1900XTX with 512 Video RAM. Sure, it is bottlenecked but I still have much love for my new card and WAS forced to buy a new one when my Nvidia blew.I maintain the 15-17FPS in major cities and airports (I run my benchmarks at Seatac), and 30+ FPS outside such areas with the following settings:*1152x864*AA Quality and AF 8x as set by my video card drivers, NOT FSX.*Frame rates set to unlimited*Water Effects Low 2.x*Aircraft Shadows Enabled*Level of Detail Radius set to Medium*Mesh Complexity 32M*Texture Resolution 2M*Autogen to Normal*Scenery Complexity to Very Dense*Scenery Ground Shadows OFFI could probably go more in depth in my FSX settings, but I think that sums it up. Why am I competing with all these high end processors? Simple, great video card and I managed to squeeze out a few MHz from my CPU. This isn't a "I got more cores" issue, this comes down to CPU raw power.[Edit: Updated cache size. I totally neglected the first time around. Sorry.)

Great idea, Kev! This cache business sure has grabbed our attention lately.Processor : AMD 64 x2 3800+ @2.0GHz (i.e. not o/c'd)L1 Data Cache : 64 KBytes x2L1 Instruction Cache : 64 KBytes x2L2 Cache : 512k KBytes x2FPS : 20 - locked and maintained 50% of the timeFiber setting : 0.50MikeEdit: BTW, you can get the CPU info from Sandra.

Could this this theory explan why some people are reporting good performance and few blurries with very low Fiber settings (e.g. 0.01)? I'm just guessing here but if I understand your theory correctly, a low fiber setting might mean the cache doesn't get emptied quite so much since the main thread gets more time and the texture fiber can only grab whats left. If your processor is fast enough, it might complete its main work before the texture fiber steps in and causes the cache to get cleared. Could people with 1Mb cache try playing around with this setting to see if a really low setting (0.01) speeds things up (while still maintaing sharp scenery)? The fiber config setting might not be related to the fibers you are talking about at all but it might be worth a try.

My fiber frame is set at .20 and I don't experience many blurries... but then again I also fly low and slow. .10 was giving me issues, my theory was to set it between .33 and .10, which led to my conclusion that .20 would work for me, and it has :)

i'm all for doing comparisons - especially as i've got Athlon 4000+ 1MB cache too.but shouldn't we come up with a standard flight to benchmark against ?i.e. take off from Airport X on runway Y, in stock airplane Z flying to Q at approx altitude ABCD. Somewhere that takes you over some high detail areas and low.only reason i say this is...when trying out new aircraft/fixing panels etc, i load up at Alice Springs in Oz since its quick to load and frame rates are 30ish since theres not much but desert.Taking off from Machiques in Venezuela on the other hand, where trees are practically overhanging the airstrip results in 15 FPS. Even with autogen tweaks (which start to make it look deserty rather than jungle [didnt change density but there's no trees for miles from the airstrip]), it's still quite slow.So my point is, if we want to compare our system specs + FSX settings against FPS, results are going to be totally different according to location.I think it'd be quite cool to have a standard test that everyone follows. At the very least, we'll try flying from a different area than where u normaly do.anyone agree, or am i talking xxxx ?If so, i nominate Kev (as OP) to come up with a little flight plan.

Agree we need a standard test. Or this means nothing.It seems like Kev is right my old AMD 3800 is smoking some of the newer cpus in FSX.I guess this is what has puzzled me when I'm doing pretty well in FSX and somebody comes along with a beast of a computer and says I get 8 fps. This may be why. If so hope they can and will patch it for you guy's that have the better cpus.

Hi Kev:Great thread!I really like the way you're carefully thinking these things out, and I look forward to learning more on these matters.I also hope you go the final distance to completion on your memory cacheing web pages, I found them quite informative, and a very interesting read!:-)BTW, if you're in the mood for a brief pause for some comedy relief, I may have come up with something which could solve all our FSX problems by taking our copmputers "Back To The Future" in a virtual kind of way .:-lol Take a break for a quick laugh like the rest of us at the EFFC forum do now and then when all this techie stuff wears our brains down... see:http://forums.simflight.com/viewtopic.php?t=55561&start=15Her's hoping to read more of your insights on the belly of the "FSX beast" when you get a chance!:-rollGaryGB

Could someone come up with a low fps scenario and viewpoint (e.g. looking down the runway towards Seattle at KSEA from the VC in an Airbus), then save the flight in this position and upload the flight plus fsx.cfg files (and possibly the config data as saved from inside FSX)? This should give at least a ballpark FPS guide without even moving. I appreciate that sampling over a 10 minute flight might give better results in the long run but, as a starting point, having the same config files and loading the flight file should give you a fairly consistent picture across different machines.I will do this myself later on if no one beats me to it. I'm not at my FSX PC right now though so it will need to wait...

*personally*, i think actual flight is necessary - as *my* frame rates seem to drop when:a) turning,:( flipping from one view to another (i.e. you were flying across barren desert, change view looking back at all sorts of detailed terrain/mountains behind which fsx seems to freak out having to do a bit more work all of a sudden.I'm sure FSX is doing a lot more work once in the air.i'm not suggesting getting all the variables the same - i.e. sharing fsx.cfg file, but just ensuring we're all comparing our hardware+fsx setup against the exact same terrain.If we then post hardware details and what fsx settings/tweaks have worked better in that same locale, relate experiences such as "flying over city XYZ brought frame rate to 2fps wheras it was 20fps at takeoff/rest of flight"... people can get a better feel for what tweaks may work better on hardware they share with others.feel like i am all talk , no action here...FSX is my first flightsim so not sure best way/areas to define a flight myself.should also point out we should all use same aicraft+panel for this experimentation, since i've noticed big FPS differences with flying say the ordinary looking default Dash8-100 compared to the lovely looking Dash8-300 british airways.

What I'm interested in is what happens when a user tweaks there own system relative to some given flight setting, in our particular case, the FIBER_FRAME_TIME_FRACTION setting with respect to level of cache. As I suggest in the readme.txt file, you could fly straight out on heading 340 since that means you will get a good mix of terrain and water.Here's my first shot at it anyway, I'm guessing we will need to refine this reference config process a bit but we might as well start somewhere...

Okay, a generic test flight. I wasn't actually trying to make this experiment into a real solid benchmark, but it seems I'm not completely off my rocker and thus would warrant a proper test.The Test Flight.I've done this flight probably about 300 times now since FSX came out, it's what I've been using to test new drivers and settings while trying to come up with a happy medium between performance and quality. I am assuming that many of us are using different tweaks and settings, but the one thing we all have in common is that our settings are giving us a decent flyable simulation, the only important factors are what your Fiber tweak is set to, what your FPS is, and if it's locked or not... is it maintaining that lock. What I'm going to do is monitor changes in FPS in comparison to actual FPS, so your actual settings don't really matter so much.Set time & weather to November 15th, 12pm, Heavy Snowstorm.Start of on the active runway at CYYZ (Toronto, Pearson Int'l). Start in a CRJ using the Virtual cockpit, don't pan around or change the zoom yet. Let the initial FPS settle, then write it down. Now look out the right window and look at the terminal building. How long does it take to render? Is it immediate? Does it stutter? Record FPS once the building is fully textured. Make sure De-icing is on.Okay now take off. Immediately after your airborne, make a hard right and center yourself on the CN Tower in the distance (it's quite a long turn), level off about 1000-1500ft and try to maintain around 230 knots. Note the distance to downtown, it's a few minutes away. About the halfway point, please take a screen shot and post it along with your reply. Monitor your FPS until you fly directly over the tower and out into Lake Ontario, keep taking screenies every once in a while to show FPS and where that occurred. Wait at least 20-30 secs between shots. Don't stop until you're screen is 100% water and snowfall.Hopefully this test will give us an indicator. One person has mentioned that maybe the Fibers I'm thinking about are not the fibers I've been reading up on. I found out by reading part of FSInsider.com, about how FSX renders its world, and of course it mentions their usage of Win32 Fibers. Another source has it that using fibers is not a good thing when sheer throughput is required by the processor, because as these processors require their large cache to maintain their efficiency, the fibers rob that from them. I'll post my test results shortly.P.S. make sure to heed the image posting guidlines of the forum. We don't need to see a 1280x768 shot in full, it can be resized to fit, as long as the FPS is still readable :)

What I think you really need to do is get someone with an Intel E6600 or better to test these findingsI dont know enough about the different Kernal Software to give an opinion - BUT I notice that so far you are all using AMD's and there IS a difference in the way Intel has worked the Conroe with it's 4Mb CacheJust an observation

Booooooo! I have to work late. I'll get my own screenies up later tonight.I should also note, that the default runway at CYYZ should be 33R. Several times I've tried it in the past and it put me on 05.... so make sure it puts you on 33R or select that manually.I will record the whole flight so people can see what I'm doing in order to recreate the flight. By the way, maintain 1500ft, go to the left of the tower, as it is actually much higher and will put a dent on your forehead if you try to fly through it :) I should have mentioned this before, apologies.

A new e6600 system arrived yesterday and let me tell you, the difference in performance is very noticeable coming from an athlon 3000+. I lock at 24 everywhere with normal autogen and high settings (global max,medium radius, 100, 19m, 1m, 2.xlow, max complexity, normal autogen, 110 mile clouds full coverage, traffic all set on 30). My bottleneck is memory, I have 1 gig, 2 more arriving soon though as FS eats all my memory (less than 50mb left) then it's HD access constantly causing stutters etc....Over Denver or seattle I'm remain locked at 24 without a dip (didn't try unlimited), until the memory runs out then all hec breaks loose. Flying outside the major cities I fluctuate between 40 and 75 with normal autogen and around 30 to 60 with max autogen, though stuttery with max autogen due to (I hope) ram limitation.Vid card is 7900 GS 256 ram. Again, my only bottleneck is ram at this point and I the fps seem very good over seattle and denver but the lack of memory stutters me up after about 5 minutes of flight as all the autogen loads into memory. I'm hoping 2 more gigs should alleviate that and I'll be able to maintain a smooth 24.I tried bloom and not much effect during the day, but at night it knocks off about 10 fps bringing me down from 24 to about 14-15. Traffic can also be a killer (default anyway) but that may be a memory issue as well.Oh, and I have the fiber tweak set at .25.Ian.