For those of you who have been wondering what FSX performance difference 1, 2 or 4 cores make in FSX, Vic (vgbaron) and I have been feverishly benchmarking away over the last week to find out.The core benchmark we used was a 5 minute autopilot flight in the default CRJ-700 that starts out a few miles North-West of downtown Seattle, over flies the city, Boeing Field, KSEA and then heads out towards Mt Rainier. The benchmark covers a majority of the scenery that FSX produces, including AI of all types, city/urban textures, water and clouds. FSX settings for the benchmark were global high defaults, unlimited frame rates, in-game AA/AF enabled, 1280x1024 res windowed, stock AI. We used Fraps to start the benchmark (P key both unpaused sim and started Fraps recording), record FPS each second, then report min/max/avg FPS at the end. Core selection was initially done using Windows Task Manager affinity settings, however we found that it produced major discrepancies in results in Vista (eg. single core turned into a blurry mess and FPS went unrealistically high). As a result, we settled on using AffinityMask in the FSX.CFG file, which produced much more consistent (and believable) results.Graphs and associated commentary / analysis is as follows: http://forums.avsim.net/user_files/177449.jpgSystem Baselines in XP - this is where Vic's and my similar systems go head to head, primarily to make sure the benchmark was providing consistent results between systems. The result showed very similar performance between our two systems, with mine having the slight edge This was probably because of the slightly more FPS friendly textures/clouds that X Graphics uses that I did not want to uninstall, so overall I'd say we have a tie at the same clock speed.http://forums.avsim.net/user_files/177450.jpgOverclocking CPU in XP - as has been covered many times in this forum, CPU speed is the primary determinant of FS performance, and Vic's overclocking efforts clearly show this here. The 46% overclock here returns and approximate 31% boost in average FPS - not too shabby!http://forums.avsim.net/user_files/177452.jpg4, 2 and 1 Cores in Vista - now we are getting into the meat of this showdown! As can be seen, there is not much difference performance-wise between running 4, 2 or 1 core. Vic initially reported a noticeable increase in blurries as the number of cores used decreased, however we now believe this was most likely due to us initially using Windows Task Manager to control core selection, which we later proved to be a flawed method of core control, particularly in Vista. My own runs of 2 or 1 core revealed negligible, if not no, impact on increasing blurries as less cores were used.http://forums.avsim.net/user_files/177453.jpg4, 2 and 1 Cores in XP - a similar result in XP, except that single core showed about a 10% average FPS reduction compared to the others. Certainly, the difference between quad and dual core runs is again negligible.http://forums.avsim.net/user_files/177454.jpgDual vs Single Core Vista - these runs were done on my new Inspiron 1520 notebook (separate review coming shortly) to clearly illustrate to problem of using Windows Task Manager in this OS to control core selection. The yellow line shows that FSX gives up on textures about halfway through the benchmark and FPS climbs accordingly because the graphics engine doesn't have that much to draw! Here's a screenie to show just how bad it got at the end of the run:http://forums.avsim.net/user_files/177455.jpghttp://forums.avsim.net/user_files/177456.jpgXP vs Vista - this final chart shows that XP is indeed slightly faster (approx 13%) than Vista with EXACTLY the same hardware configuration and FSX settings.In conclusion, there is very little difference in FSX performance between running two or four cores and about a 10% penalty (XP only) for giving FSX just one core. I'll leave it for Vic to comment on his later experience with blurries with different core counts, but my own experience between two and one core shows it to be negligible. CPU overclocking does provide noticeable performance gains, but of course this has to be weighed up with the risks of doing such an activity (which IMO are pretty low). XP shows to be a faster FSX platform than Vista by about 10%, although the upcoming Vista SP1 could change all that.The one recommendation I have out of all this is that, FOR A SYSTEM BEING OPTIMISED PRIMARILY FOR FSX, whether you buy or overclock your way there, you should get the fastest clock speed dual or more core CPU that you can, with core count being a secondary consideration. Obviously this must be tempered with what other addons / applications you want to run at the same time as an FSX session, but unless you are running something that wants a whole core or more to itself (eg. tileproxy), then my experience has shown that two cores can quite happily divide up the tasks with negligible impact on FSX. Of course, four cores are better to distributing CPU workload, but if that means a lower clock speed then FSX performance is going to suffer anyway.Whew - I'm done!Gary

Hey thanks, I appreciate you and Vic doing this for the rest of us on-the-fencers lol!A few questions:The point on your graphs where all the cpu's take a hit...is that entering SEA airport boundaries? (time 3:45ish)Also, how well did the CPUs load the scenery/textures? IE - blurries, or ok?Did you guys experiment with locking FPS and blurries/texture loading with 1,2, and 4 cores at all?How fast was the CRJ traveling?Finally, when you say settings were "global high defaults," what does that mean exactly? I assume each system had the same settings?Thanks so much, I will point this thread out to a friend of mine who just purchased an AMD6000 system....!

Gary and Vic, a most...EXCELLENT, EXCELLENT, effort!I had been WAITING for this type of post.This was professionally done.Kudos.You and Vic have made my life much easier for the next hardware step up, when that comes.I say thank you to the both of you on behalf of all the Hardware Forum readers!Cheers!Mitch

>The point on your graphs where all the cpu's take a hit...is>that entering SEA airport boundaries? (time 3:45ish)That's correct. >Also, how well did the CPUs load the scenery/textures? IE ->blurries, or ok?Whilst we didn't specifically test this, initial scenery loading is much faster with two cores than one. You'd have to ask Vic whether quad made it load faster again, but I'd say it does as this is one of the primary areas that SP1 addressed on the multi-core support front. Dynamic scenery loading kept up OK on my system, regardless of whether I ran one or two cores. Here's a screen shot at the end of each run showing dual and single core outputs - looks the same to me!Dual Core:http://forums.avsim.net/user_files/177472.jpgSingle Core:http://forums.avsim.net/user_files/177473.jpg>Did you guys experiment with locking FPS and blurries/texture>loading with 1,2, and 4 cores at all?No. Firstly because blurries were not an issue and secondly because, in my FSX experience locking FPS, does little more than kill 33-50% of the FPS you would otherwise achieve (although it does help smooth out stuttering with AA/AF on)!>How fast was the CRJ traveling?250kts @ 1500ft>Finally, when you say settings were "global high defaults,">what does that mean exactly? I assume each system had the>same settings?That means on each display setting tab (ie. graphics, aircraft, scenery, weather, traffic), "high" was selected under the global setting drop down list. This is the auto detect setting on my system @ 3.2GHz. Then AA/AF were enabled. This was saved as a CFG file and used by both of us for the runs. Gary

Just a footnote to my post for those that may stumble across this thread at a later time.Vic's system specs for the test were:Q6600 G0 CPUEvga 680i Rev A1 MoBoOnboard sound2G (2 x 1G)XP2-8500 DDR2 1066FSB Mushkin 996535 RAM320G 7200 HD partitioned for XP/Vista/Programs2 - 74G Raptors in RAID0500G 7200 HD for backupSATA DVD burnerEvga 8800GTS 640 PCI-EKandalf LCS case w/ built in liquid cooling850W Thermaltake power supplyXP and VistaMine were:E6400 B2 CPUGigabyte DS3 Rev 1.0 MoBo2G (2 x 1G) DDR2-800 Corsair XMS2 RAM250GB 7200 HDX-Fi XtremeGamer sound cardATA DVD burnerMSI 8800GTS 640 PCI-EAntec Lanboy Case480W Antec power supplyXP ProThe notebook was a Dell Inspiron 1520:T7300 2.0GHz CPU2G DDR2 667 RAM8600 Go GT 256M graphics160GB 5400 HD15.4" widescreen - 1440x900Onboard soundVista Home PremiumAnd a big thanks to my benchmarking partner in crime, Vic the vgbaron, who ran through the benchmark probably upwards of 30 times on both Vista and XP 'til we got things consistent and right!Gary

Great report/reference Gary.Any chance a copy could go to the library. The graphs will disappear in due course wont they?

Great work guys, really nice.Would it be possible to generate a graph of CPU load percentage vs time using the same test route?Not sure what tool would do that but maybe someone does.I have a quad at 3.2 and I often have task manager running to watch the cores being used. FSX does make use of the 3 and 4th core at times, it would be interesting to see with your test route where/when the extra cores were being used.Would also be interesting to have free flight load times compared.

I knew you all were up to something, and this is really great work that will help everyone.>XP shows to be a faster FSX>platform than Vista by about 10%, although the upcoming Vista>SP1 could change all that.>I hope Vista SP1 does improve FS performance in Vista. And/or FSX SP2 improves Vista performance.Reminds me of Win98SE have better 3D performance than later Win versions...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

Gary,Thanks for the post. Very helpfull.I am planning a system build specifically around use of Tileproxy and FSX. I have heard differing information about whether a quad core helps with Tileproxy texture loading during flight. Do you have first hand experience with this, or have heard anything definitive?Thanks,

I can confirm that the load times were faster with the quad. Also, with the exception of the Vista runs using the task manager the blurries had a consistent pattern. With 4 cores enabled there were blurries in the far distance. With 2 cores enabled they moved in a little closer but still tolerable. Single core was too close for my tastes. However, if you're a frame rate junkie, the frame rates were great as the graphs show.VicSystem:Q6600 G0 CPUEvga 680i A1 P30 BIOS2G XP2-8500 DDR2 1066FSB Mushkin 996535 RAM 5-5-4-12 2T320G 7200 HD partitioined for XP/Vista/Programs2 - 74G Raptors in RAID0500G 7200 HD for backupSATA DVD burnerEvga 8800GTS 640 PCIxKandalf LCS case w/ built in liquid cooling850W Thermaltake power supplyVisit the Virtual Pilot's Centerwww.flightadventures.comRadar Contact Supporter: http://www.jdtllc.com/

Yes they will Howard, although they will always be available from me on request. If there is enough interest, I'll put the benchmark and the results spreadsheet in the library as suggested.Gary

Outstanding process, data, and analysis. Thank you so much. I'm much happier with my choice to build a system based around the e6850 core 2.

Would it be safe to say that if you locked your FPS with your quad or with the duo, you would see less blurries?Oh and I should say, how were the stutters with multiple cores? For instance, with my SD cpu, I'll get micro stutters if I don't lock FPS...hehe but of course unlocked yields a much higher frame rate.

Thanks, Gary and Vic, for doing the research and writing it up. Both are big tasks, and the second one is often more tedious than the first.

I tried tileproxy a month or so ago, but gave up until I can at least leave my AV enabled and it matures a little more. If I were to run it now with a quad core, I would restrict FSX to either 2 or 3 cores such that tileproxy would a core or 2 to maximise its performance. Even on a dual core, I would restrict FSX to 1 core as the 10% reduction in FPS is more than offset gains acheived because autogen doesn't work when tileproxy is running (well, at least when I was using it back then).Gary