HyperThread - FSX - i7 -and YOU

[Guest Nick_N]

Alphahawk3 - thanks for the feedback, so HT off and affinitymask=15 works best for you. What area is a good test to show that for you?David Roch - What area are you using to test?Nick - can you comment on the RJTT flight I posted, does it stutter on your setup?Overall while the i7 clearly offers higher fps for me, when it stutters its more noticable, would like to solve that somehow...

Jack, I will look at it later today because I must restore the default scenery to check. The work we are doing on the default textures for Europe replace the ones in Tokyo... and Alphahawk3 is right.. thats one mother of an area for autogen, probably one of the worse in the sim, much worse than the LA basin or Paris. Its an area Aces should have made a custom city layout for to allow the volume 'look' of the area with a true Asian theme and without the massive autogen callsI have gotten back 2 more reports of AffinityMask=15 smoothing out i7 with HT off.. looks like there is definitely a pattern emergingSo far no one, not a single person has reported HT provided anything and the sim ran worse with it enabled in the threads I posted about its use.. but this is a on-going effort to gather data so we shall see where it stands in a few weeks too

[jackcnd 0]

Nick - Toyko is tough indeed, but its one of the few areas I can come up with that always stutters at some point.Maybe you or someone can suggest another test area so we all can try the same area?I can't say I see a clear difference with HT ON and AM=255 or HT OFF and AM=15.I can get a higher CPU clock at lower volts with HT off , so that should help FSX in general.I find the Task Manager CPU usage charts strange with AM= 15 or 255, where cores 2,3,4 all graphs look the same?

[Guest Nick_N]

Nick - Toyko is tough indeed, but its one of the few areas I can come up with that always stutters at some point.Maybe you or someone can suggest another test area so we all can try the same area?I can't say I see a clear difference with HT ON and AM=255 or HT OFF and AM=15.I can get a higher CPU clock at lower volts with HT off , so that should help FSX in general.I find the Task Manager CPU usage charts strange with AM= 15 or 255, where other than the 1st core all graphs look like mirror images?

What I suspect about this is FSX in-game use of CPU resources is multicore/multithread aware, however HT instruction is not multithread and I think the only reason FSX is seeing the virtual cores with i7 is because of changes to i7 in the core design. Intel has changed their design to a "Uncore" system in which the processor for the virtual thread actually exists (somewhat) with its own cache on the core.. this is seen by FSX as a real core but the reality is FSX is spawning threads for cores that do not really exist which may explain the activity and the heat with no result in the application. By the time the app figures out that the threads being spawned are not being addressed it spawns another thread to a real core, or, its just spawning threads that are never processed, a collision occurs and a stutter may ensue.Just a guess at this point but that would also explain the high number of returns on using AM=15 and correcting stuttersAlready have several reports load times are SLOWER, not faster as well.. so there are a few different stories floating around but the the overwhelming majority place HT as useless and AM=15 as the sweet spotEDIT: I know they made changes around photoscenery with SP1/SP2 and the engine/threads so it may be a good idea to find someone on i7 to run this same test over a total photoscenery based installed flight. If a difference is seen then there is the possibility the only time SMT may have use is with the DEM loader and photoscenery which Aces may have programed the app to handle differently. I suspect it probably will not make a difference though. I will check and see if I can find someone to run that test as I do not use any photoscenery products.

[Guest Nick_N]

EDITED WITH NEW SETTINGSJack.. just for giggles and something off the top of my headfull screen modeLoad your Tokyo flight.. then increase default AL and GA traffic to 59, set airport vehicle density to medium and cars to 7.. boats/ships to 25%I assume you are running mesh complex at 100% along with texture resolution and the LOD slider 100%. Mesh res 10m, Water at 2xLow and clouds with a 60mi draw distanceScenery Complexity to DENSE and autogen to VERY DENSE go back in the sim and let the flight start... ignore the stutters and pan visually around the outside of the aircraft (all the way around 2-3 times) then RELOAD the flight from the menu, unpuase the flight.. and pan around visually again then work frame lock, try locked 30 and unlimted with different AG and SC selectionssee if there is a change in what you are seeingBut I know Tokyo was a very poorly designed area for any system to deal withEDIT: I just did this flight with a QX6700 and a i7 940... QX6700 @ DDR3 1600 - 3.6GHz on the same settings as above, no AM edit = totally smooth. Much more stammer with AG at 100%, or, Scenery Complexity at VERY DENSE however great flight with the setttings I posted abovei7 @ DDR3 1467 - 3.6GHz Same result as above however I was able to increase autogen to 100% with no change. If I reduced AL/GA traffic to 35 and airport vehicles to low I could push Scenery Complexity up to very dense. Flight WAS 'overall' much smoother with HT DISABLED and AM=15 in the config but basically the same smooth flight as the QX which ran lower scenery settings than i7 but equal CPU speed were set in each system for the test.NOTE: Both towers have the same model 8800GTX in them, equally clockedI can not run this on the 965/GTX 285 tower right now to compare but those two towers having equal Vcards Both flights showed the smooth result with UNLIMTED frame lock... locking '30' produced a small microstutter in both systems depending on the viewThats another vote for AM=15 too and HT produced absolutely nothing in extra perf for the sim... ZIP!Of course that is just one test and I need to spend some good quality time doing a lot more, but that is all I have time for today.The issue in that flight is the Scenery Complexity.. as I mentioned earlier it is a very poorly designed area by Aces and the city buildings/landmarks themselves are the problem along with the gate/jetway data being processed at airports around Tokyo with SC set to VERY DENSE or higher.

[Guest Nick_N]

EDITED WITH NEW SETTINGSJack.. just for giggles and something off the top of my headfull screen modeLoad your Tokyo flight.. then increase default AL and GA traffic to 59, set airport vehicle density to medium and cars to 7.. boats/ships to 25%I assume you are running mesh complex at 100% along with texture resolution and the LOD slider 100%. Mesh res 10m, Water at 2xLow and clouds with a 60mi draw distanceScenery Complexity to DENSE and autogen to VERY DENSE go back in the sim and let the flight start... ignore the stutters and pan visually around the outside of the aircraft (all the way around 2-3 times) then RELOAD the flight from the menu, unpuase the flight.. and pan around visually again then work frame lock, try locked 30 and unlimted with different AG and SC selectionssee if there is a change in what you are seeingBut I know Tokyo was a very poorly designed area for any system to deal withEDIT: I just did this flight with a QX6700 and a i7 940... QX6700 @ DDR3 1600 - 3.6GHz on the same settings as above, no AM edit = totally smooth. Much more stammer with AG at 100%, or, Scenery Complexity at VERY DENSE however great flight with the setttings I posted abovei7 @ DDR3 1467 - 3.6GHz Same result as above however I was able to increase autogen to 100% with no change. If I reduced AL/GA traffic to 35 and airport vehicles to low I could push Scenery Complexity up to very dense. Flight WAS 'overall' much smoother with HT DISABLED and AM=15 in the config but basically the same smooth flight as the QX which ran lower scenery settings than i7 but equal CPU speed were set in each system for the test.NOTE: Both towers have the same model 8800GTX in them, equally clockedI can not run this on the 965/GTX 285 tower right now to compare but those two towers having equal Vcards Both flights showed the smooth result with UNLIMTED frame lock... locking '30' produced a small microstutter in both systems depending on the viewThats another vote for AM=15 too and HT produced absolutely nothing in extra perf for the sim... ZIP!Of course that is just one test and I need to spend some good quality time doing a lot more, but that is all I have time for today.The issue in that flight is the Scenery Complexity.. as I mentioned earlier it is a very poorly designed area by Aces and the city buildings/landmarks themselves are the problem along with the gate/jetway data being processed at airports around Tokyo with SC set to VERY DENSE or higher.

Just to add to thisVideo drivers in both cases were 180.48 and the OS in use was WindowsXP x64.. Both systems use a 3ware 4LPML SATAII PCIe card on single WD VelociRaptor.... FSX and the OS on their own drivesI also wish to note that when I applied the AM edit to the QX6700 system, the ground textures became mud as views changed and required more time to catch up, flight smoothnes was no different

[Guest RJTT]

Hi all,This is my first writing to AVSIM Forum. Living in Tokyo, enjoying FS every weekend, but never flown the area where jackcnd made a file to test, since passenger airliner are not allowed to fly over Tokyo after taking off from RJTT to north. I did not know how frame rate killer that area was.... I did not know how poorly designed scenery, either.I hope my reporting would be some help to all of you.(my rig is with generations old 7800GT which I will change in two months, I hope, but looks like upgrading current power supply unit from 550w to 750w may be needed before buying new VC....)CPU i940 @3.52GHzVideo Card nVidia 7800GTResolution 3840x1024x32 using TH2GDefault AL and GA traffic 59%AP vehicle density mediumVehicle 7% Boats/ships 25%Mesh complex 100%Texture resolution 100%LOD 100%Mesh res 10mWater 2xLowClouds 60mile draw distanceScenery Complexity DENSEAutogen VERY DENSE Aff.Msk N=15 HT OFFFPS target 30 average about 24fps relatively smooth with slight stutters(is it because of resolution + lack of video card performance ?)Aff.Msk N=15 HT ONFPS target 30 average about 24fps with minor stuttersNo Aff.Msk HT OFFFPS target 30 average about 20fps with stuttersNo Aff.Msk HT ONFPS target 30 average about 20fps stutters

[Guest]

Well, I tested some more with AM=15 (HT off) but I can't say I notice a difference with or without AM=15 (booted completely between tests). Everything looks and feels the same... Fps is the same (around 25 above KLAX with stormy weather, without AI though... but with the G2000 Cessna (to give it some extra load)), loading times too, smoothness also.I am not complaining though. With the i7 920 at a mediocre 3,33 (and 6 Gb ram) all runs fine. I specially noticed blurries are something from the past. I can even fly at mach speeds with time accel x2 and I won't get any blurries. Even at 4x or 8x autogen (max) keeps up for a short while!

[blueridgeflyer 1]

Well, I tested some more with AM=15 (HT off) but I can't say I notice a difference with or without AM=15 (booted completely between tests). Everything looks and feels the same... Fps is the same (around 25 above KLAX with stormy weather, without AI though... but with the G2000 Cessna (to give it some extra load)), loading times too, smoothness also.I am not complaining though. With the i7 920 at a mediocre 3,33 (and 6 Gb ram) all runs fine. I specially noticed blurries are something from the past. I can even fly at mach speeds with time accel x2 and I won't get any blurries. Even at 4x or 8x autogen (max) keeps up for a short while!

Throw me into the camp that isnt noticing a difference with HT OFF AM=15. But im certainly not one to complain as Ive been very pleased with the results in FSX on the i7.

[Guest Nick_N]

Thanks guys.. keep the reports coming!

[jackcnd 0]

Thanks for the tips again Nick, I can take this to another thread if you like, but I'll ask here since its based on your suggestions.Noticed something... When I set the frame rate to unlimited in RJTT test flight, I get massive black spikes and flashing textures when doing the initial fast pans around the plane.If I set the frame rate to 24 no artifacts during the pan.Anyone else notice the same or is that sign of a hardware or cfg issue?

[Guest Nick_N]

Thanks for the tips again Nick, I can take this to another thread if you like, but I'll ask here since its based on your suggestions.Noticed something... When I set the frame rate to unlimited in RJTT test flight, I get massive black spikes and flashing textures when doing the initial fast pans around the plane.If I set the frame rate to 24 no artifacts during the pan.Anyone else notice the same or is that sign of a hardware or cfg issue?

Its the saturated PCIe buss from autogen.. normal especially with SP2 since they moved to batch callsContrary to popular belief FSX with high settings in a very AG loaded area will saturate that buss very fast. Bufferpools a bit too high can cause those spikes to show as well as 'near' OOM with a 32bit OS too.In your case and from the description I would say you are merely seeing the buss saturated @ unlimited... I see that too here as well from time to time.. panning around clears it unless there is another problem

[jackcnd 0]

Stutter problem seems to be due to my Vista 64 install, 1st time I used V64 with the i7.Maybe some driver issue not sure. I cloned my XP FSX install from the old machine, installed new drivers, and guess what... smooth on the i7.

[Guest Nick_N]

Stutter problem seems to be due to my Vista 64 install, 1st time I used V64 with the i7.Maybe some driver issue not sure. I cloned my XP FSX install from the old machine, installed new drivers, and guess what... smooth on the i7.

Windows Xp x64 all the way for me,... and to tell you the truth I probably wont switch to W7 for FSX unless I see some real change in that OS. Right now its just a renamed Vista with fixes

[jackcnd 0]

The i7 XP install just broke the 10 sec barrier in SuperPi 1M as well!It's a 32bit XP Pro and the eVGA board can only recognize 2GB of ram with 32bit OSes,a new bios might up that to 2.49 but the new bios has OC issues for some.Think I'll try a XP 64 install never used that before, glad I joined Technet or all these OS's would break the bank.Anyone having FSX micro stutters on a eVGA i7 Vista 64 setup you might want to try XP,or maybe we can find what the Vista issue is. I going to try to compare all my drivers in both installs.I can tell you the difference for me is night and day when I fly smooth vs seeing the micro and major stuttering.Back on topic with XP 32bit, and HT off all is smooth, no AF mask entry needed, 4.1GHz.

[Guest Nick_N]

The i7 XP install just broke the 10 sec barrier in SuperPi 1M as well!It's a 32bit XP Pro and the eVGA board can only recognize 2GB of ram with 32bit OSes,a new bios might up that to 2.49 but the new bios has OC issues for some.Think I'll try a XP 64 install never used that before, glad I joined Technet or all these OS's would break the bank.Anyone having FSX micro stutters on a eVGA i7 Vista 64 setup you might want to try XP,or maybe we can find what the Vista issue is. I going to try to compare all my drivers in both installs.I can tell you the difference for me is night and day when I fly smooth vs seeing the micro and major stuttering.Back on topic with XP 32bit, and HT off all is smooth, no AF mask entry needed, 4.1GHz.

Use the 3GB switch and the USERVA edit in the boot.ini file for >2GB access in 32bit OS Jack.. my tuning list has the procedure in itNote than the USERVA amount may need to be lowered by 128 for the larger memory cards.. as long as no issues are seen such as OOM or graphic spikes the value is fine.. If those 2 conditions are seen then the value needs to be loweredIf you do decided to try XPx64, just a word of warning... do not apply SP2 for x64 from the web based updater... either get a XP x64 SP2 full install disk or slipstream SP2 into a disk before installing the OS. It does not upgrade the HAL correctly on some hardware when the web based installer is used.I also noticed when I took the i7 back up to 4GHz the AM=15 entry made no difference... but down @ 3.6 it did. When it did make a difference it was not like going from a hard stammer to smooth.. it just took out that last tiny amount of intermittent microstutter seen in the Tokyo flight so in that case it may have simply been the scenery complexity @ 3.6GHz. I wanted a apples to apples test between the two systems which is why I rant them both at equal speed/clocks