Something odd? FILEMON

[Gypsy Baron 10]

>>-SNIP->>Thanks for the info Chris.>>Related to these "random file searches" is the issue of FSX>looking for Sim.cfg before aircraft.cfg.>>By itself, a file search doesn't take much cpu resources. But>lump together HUNDREDS of file searches, and now you're>talking a real fps-robber. I would think anyway.>>>>Rhett I changed all my "aircraft.cfg" files to "sim.cfg" long ago when I first noticed this, uh, irregularity of FSX. Unfortunately, programs like aisort that look for "aircraft.cfg" folders will need to be dealt with :( FSX will also look for a lot of .CAB files before going on to find the correct file. And these are files that are never normally CAB's, like .mdl files. FSX is really VERY poorly optimized IMO, and could benefit by eliminating much of this inefficient file system access. Then there is the 30-minute, 1-hour, 2-hour, 3-hour, ... "re-validation"disk accesses as FSX goes through thousands of operations in the CryptoRSA directories....these definately cause a performance hit at those times and is especially noticable if you have a old DMA 133 hard drives, such as I do. Even though FSX is on my #2 drive and Windows is on the #1 drive, these disk accesses cause severe stuttering for 4-5 seconds at the 30 minute mark and 10-15 seconds at the 1 houtr and multiples thereof marks. As an old, litterally, assembly language programmer from the days when every CPU cycle "counted", I find M$'s coding to be an abysmal, inefficient patchwork of "eye candy centric" -expletive witheld- ! And don't get me started on "TrafficDataBaseBuilder"... Paul

[Mace 1,904]

>>IMHO FSX should load tiles according to the Level of Detail>Radius that the user defines before runtime.That would be an interesting option...>As the aircraft>moves forward tiles are discarded behind the plane and aquired>at the front. This is what FSX appears to be doing, when I'm looking at my F12 top down. ??>What FSX infact does is to aquire additional>tiles not directly linked to the main LOD seamingly at random.How do you figure? I haven't noticed this.I mean, if I'm flying around Manchester, Uk, and I'm in F12 view, I see the "tile" around manchester loaded...and I see a "tile" to the N, S, E, and W loaded in pretty high detail. Then further out in the cardinal directions I see less detail, and so on.If I move, slew, fly, whatever due North from Manchester, at some point, the "tile" to my North will pop into detail. I *do not* notice in FSX, that a tile way off my path, for say, Yarmouth (on the North Sea) is being loaded in. Are you seeing things like the latter? 'cause I'm not. But then I haven't been too observant.RhettAMD 3700+ (@2310 mhz), eVGA 7800GT 256 (Guru3D 93.71), ASUS A8N-E, PC Power 510 SLI, 2 GB Corsair XMS 2.5-3-3-8 (1T), WD 250 gig 7200 rpm SATA2, CoolerMaster Praetorian case

[Mace 1,904]

> I changed all my "aircraft.cfg" files to "sim.cfg" long ago> when I first noticed this, uh, irregularity of FSX.> Unfortunately, programs like aisort that look for>"aircraft.cfg"> folders will need to be dealt with :(>I'm glad you mentioned that compatibility issue. I guess I will have to keep mine named "aircraft.cfg" for now.Begs the question as to why a person would even program it to look for something that is not there (sim.cfg) before what is known to be there (aircraft.cfg). Just seems like a SIMPLE, and I mean simple optimization thing. There must have been a reason for it...but I've yet to hear it.> FSX will also look for a lot of .CAB files before going> on to find the correct file. And these are files that are> never normally CAB's, like .mdl files. > FSX is really VERY poorly optimized IMO, and could benefit> by eliminating much of this inefficient file system access.>Let's hope and pray that SP1 addresses this issue...potentially it's like getting FREE fps. And I am sure the team would love to breathe free fps into their sim. FPS for nothing = the Dire Straights song "Money for Nothing"If SP1 does not address this issue, I am going study it in more detail with filemon and some other things, and then make sure the FS team knows about it...again. You see, it was brought to their attn last fall. I think it was determined that there were bigger optimization issues to work out. Perhaps at the time, yes. Maybe after SP1 this issue can be revisited. From what they said, (and I could be reading into this wrongly) it seemed to be somewhat of a surprise to the team, that FSX was looking at japan bgl's when one was flying in the midlands.> As an old, litterally, assembly language programmer from the> days when every CPU cycle "counted", I find M$'s coding to> be an abysmal, inefficient patchwork of "eye candy centric"> -expletive witheld- !>When I first started programming, the instructors drummed into us to be efficient regarding cpu and mem use. We didn't have much to use, we had to be.But later, and I specifically remember this moment, an instructor said, "Well, we are not really as concerned about memory usage anymore, as machines are getting faster and have more memory these days."I said blah to that. I always want to be efficient. (If you know about assembly lang then you know all about that. Some of the best performing programs I've ever seen were assembler.)> And don't get me started on "TrafficDataBaseBuilder"...>Nah it's not too bad, it just has too strict limitations built in to checking every aspect of a flight plan. Other than that it's great. :)Too bad I can't modify the 45-minute thing between flights. In real life sometimes commuter planes take off 20 minutes after landing.RhettAMD 3700+ (@2310 mhz), eVGA 7800GT 256 (Guru3D 93.71), ASUS A8N-E, PC Power 510 SLI, 2 GB Corsair XMS 2.5-3-3-8 (1T), WD 250 gig 7200 rpm SATA2, CoolerMaster Praetorian case

[PhilTaylor 45]

>Then there is the 30-minute, 1-hour, 2-hour, 3-hour, ... >"re-validation"disk accesses as FSX goes through thousands >of operations in the CryptoRSA directories....these definately>cause a performance hit at those times and is especially >noticable if you have a old DMA 133 hard drives, such as>I do. Even though FSX is on my #2 drive and Windows is on>the #1 drive, these disk accesses cause severe stuttering >for 4-5 seconds at the 30 minute mark and 10-15 seconds at>the 1 houtr and multiples thereof marks.We have notifed the owner of those APIs, and they are investigating. A fix for that is not under our control, but we are trying to drive this. Estimated time is past SP1, I am not sure how far past.

[Gypsy Baron 10]

>>Then there is the 30-minute, 1-hour, 2-hour, 3-hour, ... >>"re-validation"disk accesses as FSX goes through thousands >>of operations in the CryptoRSA directories....these>definately>>cause a performance hit at those times and is especially >>noticable if you have a old DMA 133 hard drives, such as>>I do. Even though FSX is on my #2 drive and Windows is on>>the #1 drive, these disk accesses cause severe stuttering >>for 4-5 seconds at the 30 minute mark and 10-15 seconds at>>the 1 houtr and multiples thereof marks.>>We have notifed the owner of those APIs, and they are>investigating. A fix for that is not under our control, but we>are trying to drive this. Estimated time is past SP1, I am not>sure how far past.Thanks for the feedback. Paul

[cklink 0]

This was discussed back in December as well, regarding long load times. Adam said:"The FSX texture file loader is not operating as intended. We are, in fact, reading the entire file contents even when we only want a small portion of the file. We're investigating to see what went wrong and hopefully we can fix it." I'm hoping that means it's already been addressed and will be in SP1.That topic is here: http://forums.avsim.net/dcboard.php?az=sho...76193&mode=fullAs for more on the continual file searching, we discussed that in alot more detail here: (lots of reading)http://forums.avsim.net/dcboard.php?az=sho...77060&mode=fullI can only speak from my own results, but moving the files so they could be found resulted in the gain I posted in the topic. I rarely have any repeating files "not found" in filemon - and I have ZERO studders anymore. I guess it effects each system differently ?Regards'Garett

[n4gix 4,947]

>>Then there is the 30-minute, 1-hour, 2-hour, 3-hour, ... >>"re-validation"disk accesses as FSX goes through thousands >>of operations in the CryptoRSA directories....these>definately>>cause a performance hit at those times and is especially >>noticable if you have a old DMA 133 hard drives, such as>>I do. Even though FSX is on my #2 drive and Windows is on>>the #1 drive, these disk accesses cause severe stuttering >>for 4-5 seconds at the 30 minute mark and 10-15 seconds at>>the 1 houtr and multiples thereof marks.>>We have notifed the owner of those APIs, and they are>investigating. A fix for that is not under our control, but we>are trying to drive this. Estimated time is past SP1, I am not>sure how far past.Now that is quite disappointing to hear, seeing as how one of the other things that is happening when these validation checks occur is that some parts of the panel system are getting clobbered badly, as are the mouse routines. Too often, after the sim has passed the 30 minute mark, previously functioning "mouse points" are no longer reactive. It's as though the last clicked button on a gauge is "frozen" and cannot be changed... ;(

[dokelly07 0]

>I can only speak from my own results, but moving the files so>they could be found resulted in the gain I posted in the>topic. I rarely have any repeating files "not found" in>filemon - and I have ZERO studders anymore. I guess it effects>each system differently ?>>Regards'>>Garett>Can you expound on what it was you did, renaming wise?? Or just point me to the thread so I can figure it out. Anything to get this mess slightly settled would make me a bit happier. Thanks in advance for your time.Danon -

[GaryGB 10]

Hi Garrett:Could you post a tutorial on how to implement the changes to FSX you describe?Thanks!:-) GaryGB

[Mace 1,904]

It's in the thread/link he posted. :) But you do have a point, that the info could be place or presented in a more concise fashion.I won't be doing any file moving mods until after SP1. I don't want any problems installing SP1.RhettAMD 3700+ (@2310 mhz), eVGA 7800GT 256 (Guru3D 93.71), ASUS A8N-E, PC Power 510 SLI, 2 GB Corsair XMS 2.5-3-3-8 (1T), WD 250 gig 7200 rpm SATA2, CoolerMaster Praetorian case

[cklink 0]

Well maybe as ACES can't discuss what's going to be "in" SPI, how about telling us what's not as done above. Sounds easy enough, so is this one of those things? ;)The problem with doing what I did is that it will totally screw you for addon/patches etc in the future, so it needs to be addressed from MS end. Just as an example, lets pretend Adam's UT for FSX changed some textures to some simobjects stuff. Well now that you have moved all yours in the respective simobjects folder instead of having them scattered between other objects folders, the main texture or global/texture folders (and in some cases change them from bmp's to DDS's), - the autoinstaller probally would probally crater, or now you would end up with textures that will not get used, as they will be placed where the defaults were - and the ones you moved would get loaded first. A nightmare to say the least. Look at me, I'll be having to do a full reinstall just for SP1.I never made a list of all the files (lets just say there was lots), but explaining how I did it is easy enough. But this is an assbackwards way of fixing the issue, thus the whole reason I never made a tutorial as I said I would in the OP.I just think we should find out if this gets addressed first, if not we will deal with it after the patch is RTM. Maybe to ACES this is one of those things that will be muted by bigger and better perfromance fixes, or by just throwing more money in hardware at it. Regards'Garett

[PhilTaylor 45]

This is a good time to bring up something we have run into. SP1 will require a pristine FSX RTM.The delta patching technology we are using requires a known file to patch against, else the delta's are wrong.I know this is painful for those who have performed file tweaks, but it is what it is.Back up your changes, repair FSX, then install SP1.

[PhilTaylor 45]

As Adam stated back in December, this wasnt on our SP1 hit-list. He also didnt think it was a huge perf hit.I hope to be able to double-check this once we have the SP1 beta out and are in the final stretch, to make sure our other performance fixes havent moved this up in the performance profile.And I am trying to get a timeframe on the crypto/licensing fix.

[dokelly07 0]

So Phil, Clean the slate, install from scratch, apply SP1 - and go from there?I was thinking of doing that anyway, but especially if it is what is being advised.Danon - -

[PhilTaylor 45]

Yes, that is what we are advising.