Yeao i know this...i only say that if fibres are running ONLY in core0 then we should leave core0 alone in the affinity...like 254 for example

they're not running in core 0, they're running inside the fsx.exe user space. So if you want to change what core fsx.exe runs on go ahead if that works for you, but the fibres will follow along. I think the most benefit of Affinity masking is to get fsx.exe to run along side another program and making sure they don't conflict.

they're not running in core 0, they're running inside the fsx.exe user space. So if you want to change what core fsx.exe runs on go ahead if that works for you, but the fibres will follow along. I think the most benefit of Affinity masking is to get fsx.exe to run along side another program and making sure they don't conflict.

Thanks for the info....

Very subtle differences in the meanings of those things, But at the end of the day you don't want even 100 processes, or a 1000 threads, etc... in a single application. But obviously the less resources you give these things the more localized they get.

Lars,Thanks for the hard work you put into this research. The bottom line is that you saved me a couple of thousand dollars that I was going to put into a new PC to fly the PMDG-NGX when it's released, because for the most part FSX has been unflyable on my rig. With only the addition of the JobScheduler function to the FSX.cfg, I am now getting totally amazing frame rates on my dual-processor 8 core box, even in very complex sceneries like EDDF and EHAM. It's like I have a brand new machine that is eating FSX for lunch.I don't need an i7 anymore, and I am forever grateful for your contribution.Note to Mods: this thread should be pinned.Thanks again, Lars, and to your girlfriend :)Jev McKee

I read through most of this thread. I understand you did lots of testing, but I saw some backward performance.I'm running Affinitymask 84 on my i7, and that has given me very stable and fast texture loading, along with very good FPS.Using 249 as suggested, introduced microstutters and massively slower texture loading.I'm an user of FFTF=0 also, but be aware I've also ran 249 without any other tweaks, and testing over my known photoscenery and still saw slower texture loading.I still think that implementing HT into FSX is a bad idea.The biggest problem is visible with FFTF=0 and a heavy addon airport, and some heavy scenery, flying in VC of PMDGs 747. I get blurries after a while. While when using 84, it stays sharp most of the time.

How much antialiasing were you running? I'm pretty sure at 8xS any of the newer cards can handle clouds just fine. Even a GTX260 could suffice with locked frames IMO

I only used the in game tick box.

I read through most of this thread. I understand you did lots of testing, but I saw some backward performance.I'm running Affinitymask 84 on my i7, and that has given me very stable and fast texture loading, along with very good FPS.Using 249 as suggested, introduced microstutters and massively slower texture loading.I'm an user of FFTF=0 also, but be aware I've also ran 249 without any other tweaks, and testing over my known photoscenery and still saw slower texture loading.I still think that implementing HT into FSX is a bad idea.The biggest problem is visible with FFTF=0 and a heavy addon airport, and some heavy scenery, flying in VC of PMDGs 747. I get blurries after a while. While when using 84, it stays sharp most of the time.

FSX is a really complex program. As you say, I've done a lot of testing but it will cover far from all cases. It sounds strange that you get worse texture loading having more t&t loaders. It sounds like you have somthing running on CPU core#0 or #1 blocking the Main Thread when you use Affinitymask 249.Try with affinitymask 244. That should leave FSX running the same on Core#0 - core#3 as with affinitymask 84 on your system but putting a t&t loader on each of core#4 - core#7. Does that improve things?

Try with affinitymask 244. That should leave FSX running the same on Core#0 - core#3 as with affinitymask 84 on your system but putting a t&t loader on each of core#4 - core#7. Does that improve things?

Gives me more stutters than 84. Also more problems with texture loading.I was wondering also that you said HT is making things better, since everyone ever said to turn off HT. And I've often tested more combinations to see what impacts fluidity most.

Gives me more stutters than 84. Also more problems with texture loading.I was wondering also that you said HT is making things better, since everyone ever said to turn off HT. And I've often tested more combinations to see what impacts fluidity most.

Still strange to hear that you get worse texture loading with more T&t loaders.What happens if you run your system with affinitimask 196 and/or affinity mask 52? I can assure you that FSX can actually make good use of HT. Especially noticable if you have a dual core processor and try with and without.With a quad core the only thing you can gain by turning on HT is texture loading and load times but you will get lower FPS for overy T&t loader you add.

For the OP:Interesting angle on affinitymask settings. All I know is what I been able to observe, which is this: If I run my HT enabled i7 860 with AM=84, core 0 bounces between 10% to 30% activity, core 2 is mostly 100%, cores 4 & 6 bouce between 35% to 100, while core 1, 3, 5 & 7 show essentially 0%. If I run with AM=241, core 0 is mostly 100% activity, core 2 shows about 10% to 30% activity, cores 4 to 7 bouce between 35% to 100%, while core 1& 3 show essentially 0%. I tried shutting down my trackIR, REX weather and saitek controller to see if that effected that 10% to 30% stuff, but it had no effect; and when FSX exited that core went pretty much dead quiet. So conjecture would identify that activity as the elusive 'fibers', but really that is still a 50-50 guess. I like your idea of HT'ing the terrain/texture loading theads though, and at least for me it seems a better overall experience. I chose AM=241 since that mysterious 10% to 30% stuff won't be banging into a T/T thread sharing its phyiscal core as it does with AM=249, and it seemed to perform smoother for my setup. I think I'll keep this setting; up to now I've dabbled with AM= but was never satisfied with the results so I didn't both using it. BTW, your view on FSX 'fiber' execution is at odds with this site's resident FSX tweak guru, though I'm not sure he's right either. If you read carefully through this thread, you did somewhat contradict yourself on their execution behavior. Anyways..thanks for your observations.

Hi TheFamilyMan,The 10-30% load you see on core#0 with affinitimask 84 and on core #2 with affinitymask 241 is what I dicided to call Fibres. I'm not shure if I am actually using the correct technical term. But that is what I have decided to call them.You could also say that they are an 50% extension off the main thread that can be run on other cores (somtimes on several other cores at the same time).If you for example get limited by the GPU and fly towards a less CPU intensive area you will first see a reduction in the 'Fibres' core towards 0% and then a reduction in the main thread. During this reduction you will see small 1s 'peaks' on the fibre core every 20-35s. These 'peaks' can still be seen on the main thread once the fibre core has gone down to 0% load. (Just to complicate things futher you might actually get the reduction on the main thread first but followed by an increase on the main thread once the 'fibres' reduces). If you instead analyze the FPS graphs you will see these 'peaks' as dipps in the FPS as long as you are limited by the processor. When we get limited by the GPU the FPS graph will remain steady but you can now see these peaks on the cpu load instead. At the dipp of the FPS graph the FPS seem to be the same regardles of how many t&t loaders you have. But inbetween the dipps you have a higher FPS with less t&t loaders. One theory I have is that this is the actual fibre system we see at work with the dips in the FPS beeing when FSX is needing CPU time for the actual fibre system. (dynamic allocating of FFTF)I have not been able to see any hard evidence on any impact on FPS or anything else by letting the fibres share a physical core with a t&t loader. I was really expecting to but I have not been able to find any. I know it feels wierd.I take it you are running with an FPS lock? Otherwise your FPS is getting limited by the GPU given that the main thread is 100% load and the Fibres is not at 50% load where they should be when you are cpu limited.

This is a great thread! Original post is awesome! When I get back to my PC I will start playing around with affinity masks.. this is something I have yet to try out. I am running a Q9450, what setting would be the best for that specific CPU? Any advise would be great! Thanks!

Saab: Why whould you decide to call something fibres when you don't even know Microsoft's definition? The sad part is.. fibres are non assignable to other CPUs...and affinity masks do absolutely nothing to change that.

veeray,I decided to call them fibres as that is what those processes has been called in several other threads.cheers

no biggy.. they're commonly misused around these parts... you even get the odd person passing along Unix terms like it was 1975.