Gary,OK, I'll definitely do a quad core FSX + addon with assigned cores set of runs. One thing that would be good to do is log core usage during these runs, as FRAPS does for FPS logging. Do you, or anyone else, know of a utility that can do this?Re TH2G: there is no load on the CPU explicitly attributable to TH2G, because it is not acting as a grapics card as such, rather it acts as a monitor. All the TH2G box does is tell the video card there is a display attached capable of up to 3840x1024. Re my statement "the TH2G reduction is about 19% itself. This is pretty good IMO, as if FSX were GPU bound it could potentially be a 66% reduction": Going from 1280x1024 WideViewAspect=false to 1280x1024 WideViewAspect=true resulted in a 9% reduction in FPS due to the extra scenery and autogen being displayed in this mode. Going from 1280x1024 WideViewAspect=false to 3840x1024 WideViewAspect=true was a 29% FPS reduction. Therefore, going from 1280x1024 WideViewAspect=true to 3840x1024 WideViewAspect=true equated to the difference between the previously two mentioned comparisons which is 29% - 9% = 20% (yes I know I said 19% - sorry I goofed there :-)).The 66% potential reduction comes from having three times the number of pixels to draw if the game were GPU bound. ie. if you are tripling the pixels drawn, and the total pixels per second is pretty constant when a game is truly GPU bound, then only one third (66% reduction) of the FPS will be achieved. The fact that with FSX when you triple the pixels drawn you are only penalised 20% FPS, not the potential 66% it could be if FSX were GPU bound, which means FSX is primarily CPU bound. Hope that makes sense!I'll fire up Crysis, a very GPU bound game, today and see what impact tripling the pixels has with it - devastating I expect :-)Edit: Just did a quick check on Crysis DX9 mode, high settings, going from 1280x1024 to 3840x1024 and FPS went from 30 down to 13, a reduction of 57%. Definitely heavy on the GPU this game!Edit2: If I want to be mathematically correct on the SP1 Wide versus SP1 TH2G performance difference, it is actually the ratio from 21.993 to 17.29 = 21.4%. Regardless of specific accuracy, 21% in FSX is still a much better drop off than the 57% I get in Crysis.Gary

For those following this thread interested in FSX multi-core utilisation, check out the following charts I made of four cores slicing through FSX action:Edit - the Y axis lables are misleading, as only the Q6600 SP2 DX9 4 Core line is FPS. The rest are out of 100%, with the second chart being a stacked one (ie. 4 cores = 400%).http://forums.avsim.net/user_files/180862.jpgBiggest thing to note here is all four cores get about a 10 second flogging every minute or so.http://forums.avsim.net/user_files/180863.jpgThis shows the role that each core plays, with core 0 running flat out for the most part (that brief dip is a repeatable 0.5 sec pause just as the aircraft banks left right over the SEATAC VOR - dunno why that's happening).Before getting too excited, note that with the SP1 runs I did, dual core average FPS was only 3% behind the same run with a quad. This extra processing is likely preloaded scenery and likely translates into less stutters, but I haven't worked out how to measure that :-)I hope to have the RTM SP1 and SP2 DX9 runs completed tonight.Gary

its expected that core usage rise every minute, this is the fingerprint of the terrain texture relighting pass which happens every minute.the advice I have been giving about cores is that more cores should have some small FPS benefit but the main effect of more cores is less stutters. so this basically confirms that advice. good to see :-).I am especially curious to see your Accel/SP2 perf numbers. great stuff!

Thanks for the interest Phil. I've just finished up the Multi-Core runs with SP2 DX9, so here's the results with 1, 2 and 4 Cores being utilised by FSX:http://forums.avsim.net/user_files/180864.jpgNote the spike at one minute intervals and how with 4 cores, the peak lasts for about 15 seconds, for 3 cores (not shown due to chart clutter) it's about 30 seconds, for 2 cores it is about 45 seconds and with 1 core it never stops peaking :-)http://forums.avsim.net/user_files/180866.jpgHere is the resultant FPS and the drop is only 1.6 FPS going from 4 to 1 core. Whilst this is sort of good news, the bad news is that single core users will likely have to give up CPU cycles for addons, dual core should be pretty well optimal even with addons and quad core likely holds the smoothness crown and has horsepower to spare. I'll test the addon factor in part two of the multi-core runs.Gary

Nice work Gary. So there is value in quad vs dual. :-). I think I might point Intel to this thread.What is the percent FPS improvement in SP2 over SP1 and RTM?

This thread sure needs a sticky so it doesn't disappear. Gary, the info you are turning out is amazing. Thanks fotr doing this.Bruce.

Thanks for this Gary,Is there a way we could standardize this and make it an official FSX bench? That way various users could add their info as well..."FSX Mark 07" ;)

Yes Ryan, that is the grand plan. :-) I am working on instructions as I go, and will be able to include my standardised results as well.Gary

>>1. In addition to the frame rate differences between RTM, SP1,>and SP2, it would be interesting to know if you are seeing any>change in image quality.>There a small difference in autogen between RTM and SP1 at any given autogen setting. The reduction is more noticeable the greater the distance, so perhaps its a reduction of LOD radius and not autogen its self. The shots below were all taken using dense autogen, pre and post SP1, with LOD radius at max. I repeated this test three time to eliminate autogen variance. Pre-SP1http://sio.midco.net/FTP5/pre-sp1b.jpgPost-SP1http://sio.midco.net/FTP5/post-sp1b.jpgPre-SP1http://sio.midco.net/FTP5/pre-sp1c.jpgPost-SP1http://sio.midco.net/FTP5/post-sp1c.jpg[/img]

Gary,1. Glad to hear that you are not getting reduced ground texture clarity from RTM to SP1 as others have reported.2. 34% is definitely worth turning up the heat for. 3.6 ghz out of a quad is respectable. Are you able to do that with a stock air cooler? What CPU temps are you seeing?3. It'd be interesting to see what you get with both the overclock and the TH2GO.4. Looking forward to the seeing what happens after you startup the addons.This is great to see some of the forum theories documented. Hope you are enjoying doing this.Thanks again,Ted

Gary,You've been busy while I've been at work today.:D Very interesting. This is the kind of information I was hoping to get at the FANCON conference.I just noticed the dual boot in your sig. If you are still taking suggestions and you run out of things to test, it would be interesting to see the difference between DX9 and DX10 in Vista as well as the difference between Vista and XP in DX9.One other thing, don't let Phil talk you into being a Beta tester until you've released your benchmark utility. I'd hate to see this information get buried behind the NDA black curtain. :-hah The engineer in me is really appreciating your work here.Ted

Given I have already posted that I am tuning up Intel as to the contents of this thread, I am surprised you would suggest I would suppress this information?

Well, after many hours of running benchmarks today, it's time to finish up for the evening. I now have a spreadsheet FULL of data to analyse, which I will do over the next day or so.Two items have stuck out in all the benchmark runs that I have done that I would like to resolve before publishing the results, namely that:1. With the sole exception of the full slider right runs, DX10 is proving to be consistently slower than DX9 by quite a wide margin. I have tried two different video drivers, 163.75 and 169.09 with the same result. The attached chart shows this anomaly (see the cyan coloured line):http://forums.avsim.net/user_files/180871.jpg2. In some specific and repeatable (I did this a couple of times in each config) instances, 2 cores are proving to be faster than 4! In short, if I set AffinityMask to use Cores 2 and 3 only, FPS goes up by about 5% over quad core, but if I use Cores 0 and 1 only, FPS goes down by about 5% over quad core. The attached chart shows this anomaly:http://forums.avsim.net/user_files/180872.jpgWhilst the second item is more a curiousity than an issue, the first item definitely needs looking into. Does anyone have any idea why my DX10 results are showing lower and can suggest something I can tweak on my system to bring it up?Gary

>I worked out the flashing textures issue. Vista 64 doesn't>like dual monitor enabled when in full screen :-(>>GaryHot fix for flashing textures on multiple monitors...http://blogs.msdn.com/ptaylor/archive/2007...-available.aspx

EXCELLENT WORK, Gary. You're the hero now!! Respect.Bigean