HiRather than going back to an old thread I have dug up the oldquestion of multi-processor systems and FS9. So I have spent my40 minutes computer allowance on taking pictures of FS9 with differentsenarios while viewing the system processor + various RAM usageetc.This is for a Dual 2.8 intel with hyperthreading1G RAM800 Mhz FSBATI X600 with hypermemory + 256 Mb memoryUT-EuropeGE pro (summer textures)FS terrain 76 metre meshMax on terrainHigh on planeno AutogenSpecial effects max + add on (airports) on high.3D clouds at 100%This first picture shows a slow amble across the lake district inthe UK. The plane is the default cesna going really slow. Yousee both CPU are busy, one more so. This scene is easy to render,things are slow and the terrain is going past slowely and I get 155fps. Ok - so what !!! There is 450 Meg of physical memory free,and little has been paged.Now I up the speed and the f14 is just going through the soundbarrier over London at dusk with the UT lighting model on andCity airport (plus downloaded addon) in the distance. This is heavyon data thoughput as the scenery is rushing fast at about 700 knots,the scenery is busy with extra lighting model from GE pro (though a bit light to see at the moment) + there is free ware addon airport.Now we see both CPU hitting 90% occasionally, occasional blurringof textures as the data transfering gets overloaded. Still nodata paged, with 700Mbyte of physical memory in use. At this pointthe limiting factor is data throughput, and this probably preventsCPU usage getting higher, and this CPU usage is quite variable dueto data transfer waits. This time 40 fpsThis time I use the bluesky scenery in LAX, so now we have the addon and special effects on max to cause problems, so the limitingfactor is CPU here with matrix multiplication of objects. thistime we see both CPU consistently over 80% and peaking at 90%. Noticethat this is still the same FS9 session, but there is no pages ondisk and 700Mbyte of physical memory in use. Still no pagingcausing problems . fps about 20 - due to the high qualitysetting of sceneryFinally, here is a finished flight screen, and we see 1 CPU approachidle, though 1 is till busy (? congrats MS).When FS9 is finished both CPU hit 0% (ie less than 0.5%), the datastays resident in memory. You can see no effect of the antivius,firewall, etc.Conclusions - open to comment.Win XP is able to apply a threaded process to multiple CPU attainingup to 180% over a single processor for FS9. I don't get this closeon other programs, but still see both CPU being used. I have tunedFS9, other programs tend not to be so aggressive.1Gbyte is enough even with UT, GE pro, 76 metre terrain, no paginghappens within this session of about 40 minutes. Of course if youfly hours there will eventually be paging. So I don't need more memory for FS9 on win XP. Of course, FSX and Vista is another issue, and you will need to double memory for "each".You can get good fps even with a modest system, and there is spacefor other programs to be running. I see no CPU usage from myantivirus & fire wall. I don't get stutter even when I push things to the limit if youset parameters right. fps is good. The terrain gets a bit mushed at 800 knots, but I am not looking at the scenery at thisspeed !I have not set the view for the 4 virtual processors in thiscase, not that important, though I guess it might give a clueon 4 CPU.RegardsTom

PS These were done with windowing, I get similar results with fullscreen, but the pictures were harder to get. Full screen justputs more load onto the graphics card with higher resolution,AF and AA etc. Does not effect the figures above.

Thanks TomOO, for your elaborate test/comparison. I thought that dual core wasn't very helpful with FS9.1 but you seem to prove otherwise.Rob "Holland&Holland" de Vries http://fool.exler.ru/sm/fly2.gif"To go up, pull the stick back. To go down, pull the stick back harder"

I get much different results.

Comments ...What I believe (though I am a linux person not a win person), isthat you are looking at FS9.exe as a specific process. This process analysis is likely not to look at threads generated byFS9 - though maybe someone can comment on this. Because of this, your FS9 process is not actually working veryhard with a minimal CPU usage. FS9 consists of many dll (dynamicload libraries), and if these are spawned as a threads they do notregister under the parent process when you look at CPU usage forFS9.exe (They certainly would not under linux). Interesting,the memory IS registered under the parent process - again thisis the same as linux.If you look at FS9.exe it is actually quite small in size, and only does licensing, and control of the main flight process. In fact, even if you were to look at all the parent (FS9.EXE) andits child processes (ie spawned threads), you would still not seeall the CPU time. This is because FS9 (and FSX) are very aggressiveprograms put strain on Disk IO, and data transfer - and these arehandled by the windows kernel processes - ie show up as separatehigh usage applications as they will be serving the simulation butnot actually part of the simulationI specifically looked at the total CPU usage of my system andcompared this with a resting state. In fact I have seen only 1cpu used when on the start up screen, but the second quicklycomes into play when flying.Hope this explains.RegardsTom