This morning I thought I might benchmark the difference between no affinity mask and affinity mask=15 (all four cores) on my quad using a 150sec autopiloted, locked spot external view flight.My PC specs are in my sig but main fsx settings are as follows:Res: 1680x1050x32Target Frames: UnlimitedGlobal Texture Res: Very HighLevel of Detail Radius: MediumMesh Complexity: 56Mesh Res: 38mTexture res: 1mWater effects: Low 2.xScenery: NormalAutogen: SparseClouds at 60miles and High Density CoverageAirline Traffic: 15%General Aviation Traffic: 15%No Land and Sea TrafficI am also running FS Genesis Mesh and the Cloud9 XClass landclass for Australia. No Active Sky for these benchmarks as it is too unpredictable between loads (and we are in storm mode here so the plane keeps overstressing when it hits severe wind gusts).The chart below tells it all. BTW, the two sudden drops in FPS occurred when the terrain shifted and the camera had to readjust signficantly to maintain its viewpoint - the plane just got there quicker in one test.Andrewhttp://forums.avsim.net/user_files/174756.jpg

And added affinity mask=3 (two cores) and a no affinity mask but wideview=true tests (the latter definitely cuts back on FPS).Average FPS per test (150 seconds) work out as:Affinity Mask=15 - Mean Average of 59.60FPS, Median of 61FPSNo Affinity Mask - Mean Average of 59.28FPS, Median of 60FPSNo Affinity Mask & Wideview - Mean Average of 55.57FPS, Median of 57FPSAffinity Mask=3 - Mean Average of 58.89FPS, Median of 60FPSAndrewhttp://forums.avsim.net/user_files/174757.jpg

Just goes to show those extra cores seem to do a whole lot of nothing in the FPS stakes with FSX. Pretty disappointing really, considering my second core gets taken away from doing something noticably constructive, like handling OS processes and addon loads, while it is otherwise wheel-spinning in FSX. More often than not, I now set affinity to one core when running FSX. Gary

Yep, those poor cores don't know what's going on :-)I might try a few other tests, tweaking a few settings in the fxg.cfg, but i don't expect different Affinity Mask settings to have any effect. I would like, though, to measure any FPS impact with changed TEXTURE_BANDWIDTH_MULT and FIBER_FRAME_TIME_FRACTION settings (compared with visual improvements etc) so will give that a shot later in the week.Andrew

http://forums.simflight.com/viewtopic.php?t=62147From the post:"There is a tweak to control scheduling of threads on cores.It is not recommended to change these settings unless you have a performance reduction and/or maxed out CPU loads in the PerfMonitor. The tweak isEdit:Fixed a typo here[JOBSCHEDULER]AffinityMask=nwheren num of cores scheduled1 = 1 core 00013 = 2 cores 00117 = 3 cores 011115= 4 cores 1111"It made my second (Hyper) thread of my old P4 go like a banchee, but will it matter to a real dualie?

Well, my first core (core 0) is pretty much maxed but obviously setting affinity mask does not force much if any load onto the other cores. I had a dual core (E6600) before the quad and the results were almost identical.Andrew

Maybe Phil Taylor will see this thread and comment......

>It made my second (Hyper) thread of my old P4 go like a>banchee, but will it matter to a real dualie? The question is, with all that extra core action, what does all this effort translate to? - More FPS? - nope, in fact single core ops give a measurable few percent more FPS. - Less blurries? - no difference as far as I can tell. - Less stutters? - if anything slightly more because other apps and services are now being starved of spare core capacity.Seriously, if the other cores are being tasked to pre-process upcoming scenery or AI, wouldn't this effort EVENTUALLY translate to either higher FPS, or less blurries and stutters?Gary

I don't believe this. After weeks of seeing slightly better FSX performance with one core rather than two, since I made the above post the opposite is now true. On a flight I am currently doing in the LDS 767 from Sydney to Brunei, over the middle of the Australian outback, I was getting (as verified by running multiple 15 second FRAPS FPS averages):- Core 0 only - 23-24 FPS - Core 1 only - 25-26 FPS (always goes slightly better than Core 0, because system processes still seem to run on Core 0 even when it is maxed out by FSX)- Cores 0 and 1 - 27-28 FPS (for once, dual core gives nearly 10% advantage over single core!)The only change I have made over the last day is to knock down my sound acceleration in dxdiag down one notch to standard acceleration, which has helped reduce stutters quite a bit. If a side effect is that multi-core operations are now more effective, then I'll be more than happy!Gary

>Maybe Phil Taylor will see this thread and comment......Well I think he'd say the extra cores are not supposed to do much to fps, and these results bolster that.The threading work done in SP1 is aimed at load times...primarily...RhettAMD 3700+ (@2585 mhz), eVGA 7800GT 256 (Guru3D 93.71), ASUS A8N-E, PC Power 510 SLI, 2gb Corsair XMS 3-3-3-8 (1T), WD 150 gig 10000rpm Raptor, WD 250gig 7200rpm SATA2, Seagate 120gb 5400 rpm external HD, CoolerMaster Praetorian

Interesting. Another benchmark to run :-)Andrew

As I understood it from the blog, the terrain threads were being spawned across as many cores available. Maybe a 'stutter' comparison would be a better test for multicore performance?Although I'm not sure how you would do it.............

Gary,I find with the LDS767 that my FPS can vary from one day to the next in the same area without any explanation.For instance, last two nights I have started at KSYR in 'fair weather' with zero AI turned on. First night I was getting 22-23 FSP in the VC. Last night I was getting 18-19.FSX seems like a strange beast to me.............

>As I understood it from the blog, the terrain threads were>being spawned across as many cores available. Maybe a>'stutter' comparison would be a better test for multicore>performance?>Although I'm not sure how you would do it.............Stutters and blurries also. I am also not sure how to quantify it. There was REST_MCP / Gary's FSX benchmark thread. Maybe it would be a good flight to use.RhettAMD 3700+ (@2585 mhz), eVGA 7800GT 256 (Guru3D 93.71), ASUS A8N-E, PC Power 510 SLI, 2gb Corsair XMS 3-3-3-8 (1T), WD 150 gig 10000rpm Raptor, WD 250gig 7200rpm SATA2, Seagate 120gb 5400 rpm external HD, CoolerMaster Praetorian

>On a flight I am currently doing in the LDS 767 from Sydney to>Brunei, over the middle of the Australian outback, I wasI am doing exactly the same flight right now! Are you following the WOAI releases too ? BTW thanks for the dxdiag Sound acceleration tip, it seems to help with micro stutters during take-off. ==Marten:Weber==P-D920 oc3.64MHz 2GB RAM WinXP ATI X1900+1024MB 2xSATA250 2xVIEWSONIC 19"Need AI? Check www.world-of-ai.com!