Archived

This topic is now archived and is closed to further replies.

RESET MCP ALT

FSX Dynamic Benchmark Run Results

Recommended Posts

After running over 40 benchmark runs over the last few days on both Vista 64 and XP Pro with FSX RTM, SP1, SP2 DX9 and SP2 DX10, not to mention a dash of overclocking, widescreen, TH2G selective CPU core use and different FSX settings, I have finally cobbled together the results as shown below.First up is the test config and legend for the results:http://forums.avsim.net/user_files/180939.jpgNext is the test results themselves, limited to just the average FPS of the run (I have second by second blows of the full 5 minutes of each run and can draw up pretty graphs for any set of these runs if anyone wants them :-):http://forums.avsim.net/user_files/180940.jpgThe conclusions I have drawn from this effort are as follows (calculations not shown in the name of sanity):1. SP1 gives just over double the performance of RTM and SP2 adds another 10% with DX9 on Vista 642. SP2 DX10 is 25% slower than SP2 DX9, however when bloom is enabled, DX10 performs the same but DX9 loses 37% performance on Vista 643. Max sliders is unplayable in SP1, SP2 DX9 and SP2 DX10 at around 6 FPS average and mass blurries set in within 30 seconds on Vista 644. My SP1 optimum setting configuration goes from a 13% benefit to a 5% deficit between SP1 and SP2 on Vista 645. Enabling widescreen support causes a 9% performance drops with SP1 and only a 5% performance drop with SP2 on Vista 646. Using TH2G, which increases pixel output threefold, elicits a nearly 30% performance drop in SP1 but a lower 23% performance drop in SP2 on Vista 647. Using cores 0 and 1 out of all four cores incurs a 3% performance loss, however gives a 7% performance boost when cores 2 and 3 are used on Vista 648. A 50% CPU overclock translates to a 34% performance gain, which is an efficiency factor of 68%, on both OSs9. Running 1066MHz FSB versus 1600MHz FSB results in neglible performance difference (<0.5%) on Vista 6410. The highest recorded performance was 35 FPS average with SP2 DX9 Global High settings and the CPU overclocked to 3.6GHz on Vista 6411. XP Pro is 4% faster than Vista 64 with RTM12. Vista 64 is 4-6% faster than XP Pro with both SP1 and SP2, with CPU overclocking causing the highest difference13. The gains in XP Pro going from RTM through SP1 and SP2 are approx 10% less than with Vista 6414. Reducing FSX core use from 4 to 2 cores with XP Pro resulted in a 2% (cores 0&1) to 7% (cores 2&3) performance boostFinally, some general comments I have are:1. Image quality was very consistent between most runs, with the only noticable degradation being mild blurries when I did a single core run and major blurries during any max slider run2. I did not get to test FSX with multi-addon performance as I have not yet installed my addon suite and will need to consider how to get consistent and repeatable results between runs3. FPS is only one aspect of the FSX game experience and should be balanced against visual quality for optimal fun factor.As previously mentioned, I will package the benchmark, along with instructions, and submit to the AVSIM library sometime over the next week for others to torture themselves with ;-) Enjoy!Gary

Share this post


Link to post
Share on other sites
Help AVSIM continue to serve you!
Please donate today!

A quick footnote on FSX's apetite for cores. On some of the 2, 3 and 4 core runs I recorded FSX total CPU utilisation. Every minute the sim runs up all available cores for scenery pre-loading (I expect) and the amount of time it spends doing so is inversely proportional to the number of cores it has at its disposal (ie. less cores mean more time pre-loading).What is interesting however is that average CPU usage over the entire run across 2, 3 and 4 cores is roughly the same at around 1.5 cores of equivalent full time use. This means that a dual core has 25% spare capacity to run addons with FSX, which seems optimal, and that a quad core has horsepower to spare. Where that leaves single core, I am not sure because FPS certainly doesn't dip by 33% when a single core can only serve up 2/3rds (ie. 1 core's worth) of processing power to FSX's 1.5 core average apetite. Food for thought anyway.Gary

Share this post


Link to post
Share on other sites

That is extremely well done and answers so many questions.Thanks!

Share this post


Link to post
Share on other sites

Truely inspiring work, this will help inform my buying decisions, both hardware and software. A couple comments:>10. The highest recorded performance was 35 FPS average with>SP2 DX9 Global High settings and the CPU overclocked to 3.6GHz>on Vista 64What about the 3 core configuration that scored 37? Fascinating that fewer cores helps performance.Interesting too about the FSB. So bottom line is to get the highest CPU clock multiplier you can afford if you don't want to overclock.I guess you are bored with all this now, but I'd be interested to see how the following influence your baseline:1) Changes in RAM timings2) Under/over clocking video.Would not expect any great difference, but makes a complete picture. Certainly I will try these when your benchmark pakage is available. I have a 7600GS and X800XL, so will try both.Many thanks for taking the time to do this.

Share this post


Link to post
Share on other sites

Great work as always Gary. It confirms my reduction in performance with the DX10. At least I will finally have learnt not to jump in with major harware purchases for FS11.

Share this post


Link to post
Share on other sites

Wow, thanks for the effort and it must have been a effort indeed!Looking forward to your tool, I'd like to see my results at 1920x1200 with different configs.

Share this post


Link to post
Share on other sites

It seems I missed a few run results in the run table due to hidden columns in Excel (I was trying to reduce clutter :-)). Here is an updated chart with all test run results:http://forums.avsim.net/user_files/180948.jpgAlso, I had rejigged the order of the results from testing to group like results together, but of course that made me miss one of the logical groups that contained the 37 FPS result you mention. Here is the 37 FPS run result alongside its related runs:http://forums.avsim.net/user_files/180949.jpgThis amends Conclusion item 10 to now be:10. The highest recorded performance was 37.16 FPS average with SP2 DX9 Global High settings and the CPU overclocked to 3.6GHz and limiting CPU cores used to 2 and 3 on Vista 64Re RAM timings and video over/underclocking, I did not want to do too many hardware specific runs for fear of my post being banished to the hardware neverland forums and thus getting lost in the ether (whooooosh, here we are anyway! :-lol). Seeing that we are here now anyway, I will add a few extra hardware variations like you suggest above to my test suite in slower time and will report back here.Once I post the benchmark suite to the AVSIM library, you can always run such tests yourself anyway.Gary

Share this post


Link to post
Share on other sites

Excellent data Gary, I'm still absorbing it all, but have two questions. You listed your graphic settings for MAX and OPT, but I don't see where you mentioned what your baseline RTM, SP1, and SP2 settings were. Would you list these?Were your WIDE, TH2GO, 3.6 GHZ etc. runs done at the baseline graphic settings or the OPT settings?Thanks,Ted

Share this post


Link to post
Share on other sites

Ted,>You listed your graphic settings for MAX and OPT,>but I don't see where you mentioned what your baseline RTM,>SP1, and SP2 settings were. Would you list these?If not mentioned, then they are as defined in the baseline configuration, namely Global High defaults, Realism hard, Target frame rate unlimited and in-game AA/AF enabled>Were your WIDE, TH2GO, 3.6 GHZ etc. runs done at the baseline>graphic settings or the OPT settings?They were done at the baseline graphics settings, with the idea of keeping the FSX settings the same wherever possible between runs of a particular configuration.Gary

Share this post


Link to post
Share on other sites

Gary,Okay, cutting to the chase here and to save my infeabled sanity, let's cut to the chase... Assuming my wife gives me the needed $$$'s, what is your recommendation for the best hardware to purchase starting from the ground up? Once you have answered that question, I will invoke your name with the reverence it deserves in begging my wife to increase my allowance long enough to buy one of those whatzits. :)

Share this post


Link to post
Share on other sites

Gary,two questions:1) weather? (clouds? how many layers? rain? etc.)2) traffic. I can see it's set to 80%, but the impact of 80% in JFK or LHR is not the same as in some remote airfield.thanks for your efforts. This is quite interesting!!//Michael

Share this post


Link to post
Share on other sites

Fantastic Gary! You are the MAN!One question please,...what exactly is WIDE SCREEN MODE?How do you define that?Thanks for all the work that went into this report!!!Mitch

Share this post


Link to post
Share on other sites

Michael,Weather is based on a saved download I generated from ASX a few months ago, which is more complex than the default fair weather theme (but not overly so, and gives an interesting wind shift overflying KSEA. I wanted something a little more demanding than the default fair weather but didn't want to choke the whole graphics engine up with massive layers of clouds and rain. Traffic is based totally on FSX default as there are just too many variables when addon AI traffic comes into play (eg. # of flight plans active, complexity of AI models, time of day). Again, I didn't want to choke the benchmark with just this one aspect of the FSX experience. Nonethless, with Max settings you will see quite a lot of traffic on the ground at KSEA and a few in the air. Anyway, you'll get to see what I used when I upload the benchmark package to AVSIM. You can also run the benchmark with any weather or AI set you like, but it will obviously make it much more difficult to compare performance with other people's systems.Gary

Share this post


Link to post
Share on other sites

Mitch, In FSX.CFG there is a line that says: WideViewAspect=False, which can be changed to True to give a wider FOV at the same zoom level. Without it, FSX on widescreen, and especially TH2G, looks a bit tunnel visioned. There is of course a performance impact of FSX drawing more objects for your wider FOV, but this comes down to a little pain for a lot of graphics gain IMO.Gary

Share this post


Link to post
Share on other sites

In your FSX.cfg file, Section [Display] there should be a line called WideViewAspect and it usually is set to false. Make it read:WideViewAspect=Trueand you'll have 16:9 aspect ratio view (resolution can be anything, but it would be like watching a 4:3 tv show on a 16:9 tv - all stretched with this option set False).

Share this post


Link to post
Share on other sites

Tom,From my perspective and experience, the essence of a fast FSX system, in order of priority, is the following: CPU - two or more CPU cores and the fastest clock speed you can buy or otherwise achieve through overclocking. Intel reigns supreme in this arena at this time.CPU Cooler - if you are going to overclock, get at least one of the value aftermarket HSF combinations, like the Arctic Cooler Freezer5 Pro 7 that I use. After that, they can get very big and expensive for diminishing cooling returns.Motherboard - if you are going to overclock, get a brand name, known overclocking friendly motherboard. It doesn't need to be the Deluxe++ WiFi Limited Edition of a particular board series, as quite often the value versions of these allow just as good overclocking, just minus the bells and whistles that don't matter much to FSX performance (eg. RAID, Firewire). Newegg customer feedback comments are your friend here.Power Supply - you'll need at least 500W PSU and one of good quality. I'd much rather have only 500W of quality power than 1KW of ripply volts and amps from a dodgy brothers unit. 500W is sufficent for the base unit I describe here. FWIW, I use a 480W ANTEC with the rig I used in the tests above, with no issues at all.Case - Again, if overclocking, make sure that the case you get is well ventilated and can accommodate any huge aftermarket HSF you buy.RAM - 2GB for XP and 4GB for Vista (with a strong recommendation for Vista 64 to make full use of all 4GB). Memory speed and timings don't make a huge difference (I will be testing for this shortly), just as long as whatever you buy run at your potential overclock speed. DDR2-800 at 5-5-5-18 timings is plenty good enough for what most want to do here.Video Card - either a 7900GT or 1950XT for DX9 only, or an 8800 GT for DX9 & DX10. Although video card performance does play a greater factor in FSX than any previous FS version, particularly if you want to crank AA/AF levels up beyond what in-game setting provide, having a fast CPU and enough RAM to feed it will do much more to bring up your low-end FPS with this title. No SLI or Crossfire, unless your house needs the extra heating ;-) it is pretty much entirely useless with FSX! Edit: And get a 512MB or greater version if you run a resolution over 1280x1024 or really like to push those AA/AF levels up to max at the video card driver level.Hard Drive - I know that some swear by RAID and Raptors, but in my experience it doesn't make a cracker of difference while FSX is chugging along, as long as your FS scenery directories are regularly defragged and you are running a 7200RPM unit. When considering hard drives for an FSX system, I think much more about capacity than performance.Sound Card - Any modern, name brand, even value, overclockers motherboard comes with a more than adequate for onboard sound capability for FSX performance consideration. Don't bother with an external sound card for FSX unless you really have a fine ear for the sound quality difference it offers over onboard.That pretty well covers the major system components. Of course if you play other games, especially First Person Shooter, then I'd consider upping the video card spec to 2 x 8800GT in SLI and say a 800W PSU to handle the extra load.Finally, here are some recommendations for FS-optimised rigs (minus monitor, mouse and keyboard) I put together a couple of weeks ago:Overclockers Performance System ~$1300 for 3.2-3.6GHz quad core DX10 power - http://secure.newegg.com/NewVersion/wishli...tNumber=6548852Performance System ~$1150 for 3.0GHz dual core DX10 power - http://secure.newegg.com/NewVersion/wishli...tNumber=6551312Value Overclockers System ~$600 for 3.2-3.4GHz dual core DX9 power - http://secure.newegg.com/NewVersion/wishli...tNumber=6551492Gary

Share this post


Link to post
Share on other sites

Mitch, In FSX.CFG there is a line that says: WideViewAspect=False, which can be changed to True to give a wider FOV at the same zoom level. Without it, FSX on widescreen...----------------------------------Thanks Gary! I'll try to find the same in FS9....if there is one. I thought that was what it was, but then also thought it could be for multi-monitors...:)Mitch

Share this post


Link to post
Share on other sites

Hi Garry, Your hardware recommendations are spot on IMHO. Pretty much exactly what I thought is required for FSX (and works pretty darn well for Crysis as well). On that topic, I have to chuckle about the legion of fans moaning about the performance of Crysis; we have been doing battle with FSX for over a year and I have a pretty well sorted system now, so problems are minimal.Cheers,Noel.

Share this post


Link to post
Share on other sites

I'm looking for a couple of volunteers to beta test my benchmark package before I put it up in the AVSIM library for public consumption. Nothing more than using the instructions to do one run through the standard test and reporting results and comments back to me.Please email / PM me if you are interested in helping out and I'll email you the package.Gary

Share this post


Link to post
Share on other sites

Excellent work. I will publish fps rates for this system in the coming weeks: QuadExtreme 45nm 9650 CPU @ 4 MHz and 5 MHz / Asus P5E3 / DDR3-1600 / Asus EAH3870 DX10.1 graphics / Raptor HDs/ I am in the process of installing everything, then run a burn-in for a week, then proceed to install Vista 64 and FSX. First tests with the system in a slightly different test configuration with a different graphics card show more than double the rates of the Q6600 reported above.

Share this post


Link to post
Share on other sites

Thanks, Garry, for your huge effort and very interesting and useful information.Dirk.

Share this post


Link to post
Share on other sites

Hi Gary, great stuff.. Any chance you could test on Vista 32 bit? I'm ready to build a new rig and am just nervous about using 64 bit because of all the driver incompatibilities and problems with other software.thanks again,Kev

Share this post


Link to post
Share on other sites

Kev, New rig and driver compatibilities with Vista 64 are pretty well a thing of the past. I had the same concern but a few here in this forum talked me around, and now I'm loving Vista 64!Getting older programs to run may be a different issue, but I have spent most of this afternoon installing all my 32 bit favourites and haven't come across a single compatibility issue yet (touch wood!).Re doing 32 bit runs, since I don't have a Vista 32 installation DVD, and wouldn't want to undo all the goodness I just setup with Vista 64, I won't be able to do any runs on my rig with Vista 32. Sorry!Gary

Share this post


Link to post
Share on other sites

>QuadExtreme 45nm 9650 CPU @ 4 MHz and 5 MHz / Asus P5E3 / DDR3-1600 / Asus EAH3870 DX10.1 graphics / Raptor HDs/5 MHz and FS1 ? :D However, good luck with your 5 GHz attempt...Gary, congrats and thanks for this excellent comparison. With my 3.6 GHZ Conroe i was chewing through Vista 32/2GB and Vista 64/4GB as well and can confirm your findings. Well, that Vista 64 is that strong against XP i didn't know. Very interesting !Heiko

Share this post


Link to post
Share on other sites