FSXMark11

[cmeeks 8]

I'm pleased to announce the latest and greatest in FSX benchmarking - FSXMark11. Mad props to the developers of the original FSXMark07 from which this newer version gets its' roots. Download it, follow the guide to a tee, and come back here to report your results. If you have run FSXMark07 in the past, you will notice this newer version is much more taxing. This was done to get a better concept of realistic frame rates on todays hardware.We are all very eager to see the results from a variety of systems, be it AMD, 775, 1366, 1155, etc.Download: FSXMark11

[cmeeks 8]

I started this mess, so I'll be the first to post my results. Notable FSX.CFG Tweaks:[bufferPools]UsePools=1RejectThreshold=98304PoolSize=8388608[JobScheduler]AffinityMask=14[Graphics]HighMemFix=1Software Configuration:NVIDIA Inspector disabledDefault FSX texturesSee guide for detailsHardware Configuration:CPU: i7-2600k @ 4.8GHz HT OffRAM: 2x4GB DDR3 1866 (9-10-9-27)GPU: GTS 250Results (average of 2nd, 3rd, and 4th run):Frames: 8800Time: 300000Min: 19.3Max: 31.0Avg: 29.3

[SpiritFlyer 987]

Average of 3 Benchmark Runs: Frames 11191, Time 300000(ms), Min FPS 23 Max FPS 49 AVG FPS 37.30Run 2.Frames, Time (ms), Min, Max, Avg 11256, 300000, 24, 49, 37.520 Run 3. Frames, Time (ms), Min, Max, Avg 11284, 300000, 21, 50, 37.613Run 4. Frames, Time (ms), Min, Max, Avg 11034, 300000, 24, 47, 36.780----------------------------------------------AverageFrames, Time (ms), Min, Max, Avg 11191, 300000, 23, 49, 37.30CPU i7 2600K (see above settings)MB ASUS P8P67GPU GTX 580RAM 8g Mushkin Redline #996805 (see above settings)OS Windows 7-64 sp3Test OS Drive: SSDTest FSX Drive: Black Caviar 64 cache 640gDefault scenery, default enviroment, default textures, default mesh Adherance to all stipulated guidlines and settings as stated in official Benchmark instructions as of this time and date of posting.Major TEST fsx.cfg adjustments:[bufferpools]PoolSize=10000000RejectThreshold=131072[GRAPHICS]HIGHMEMFIX=1[JOBSCHEDULER]AffinityMask=15

[dazz 69]

Tweaks[JOBSCHEDULER]AffinityMask=14[bufferPools]RejectThreshold=131072UsePools=0ResultsFrames, Time (ms), Min, Max, Avg 13511, 300000, 27, 60, 45.037Frames, Time (ms), Min, Max, Avg 13375, 300000, 29, 58, 44.583Frames, Time (ms), Min, Max, Avg 13379, 300000, 26, 59, 44.597Frames, Time (ms), Min, Max, Avg 13481, 300000, 27, 59, 44.937...something's not right

[cmeeks 8]

[bufferPools]RejectThreshold=131072UsePools=0...something's not right

I was looking at Bojote's FSX.CFG and I noticed your RejectThreshold value was considerably higher. Then I found this post:

UsePools=0 gives you a more 'consistent' experience with no stutters on 'heavy' scenery areas, but it also 'lowers' your overall FPS if you are flying in complex scenery AND if you have the Vsync option in FSX set to on. So, the only way to take that FPS back is to 'balance' things a bit, meaning, use the BufferPools but ONLY for 'small' chunks of vertex data, so it is much more efficient and you make better use of your Video Card capabilities. This is achieved using the RejectThreshold setting. The 'optimal' setting being 'exactly' 96KB ;) any lower and you loose the benefits of the BufferPool, any higher and you make the CPU work harder. This is, of course my 'system' optimal setting. It will vary from hardware to hardware.

Considering you're running a highly clocked Sandy Bridge CPU, perhaps your higher RejectThreshold value is serving to allow the CPU to do an appropriate amount of work?

[dazz 69]

I was looking at Bojote's FSX.CFG and I noticed your RejectThreshold value was considerably higher. Then I found this post:Considering you're running a highly clocked Sandy Bridge CPUs, perhaps your higher RejectThreshold value is serving to allow the CPU to do more work.

Thanks for the HU. This is all so confusing man, that's the rejectThreshold that the venetubo app itself suggested, and I thought the UsePools=0 thingy worked the other way around, increasing performance at the cost of stability/stutteringI'll try different RT values and see what looks better

[SpiritFlyer 987]

I am at a loss to explain the wide difference between our 3 systems since this is a highly intensive CPU test and we are so close in specs. My entire system is heavily modified with FTX adjustments and enhancements that at least partially apply even when the ORBX scenery is not loaded, so that might have something to do with something, but I just don't know. Some comparative testing is in order between using my Bufferpools settings and the one used by Dazz to see if I can isolate whether ot not that makes the difference. Even that would hardly explain the huge variation between Corey's numbers and that of Dazz's I am afraid. We need some ideas here.Kind regards,

[dazz 69]

It's graphics. I got a big perf boost disabling the Inspector profile and letting FSX handle FSAA & AFAre you guys sure you did that too?Just to clarify, I mean graphic settings in mi side. The test is perfectly set up for a CPU bound scenario

Edited February 24, 2011 by dazz

[cmeeks 8]

How did you go about disabling your Nvidia Inspector profile? I assumed removing the FSX.EXE application from the MS Flight Simulator X profile would effectively disable it, so that's what I did. Also, should I have specified Anisotropic instead of Trilinear filtering in the benchmark???Stephen, while this is a rather CPU intensive test, you must remember that my GTS 250 only has 128 shader cores. Compare that to the 336 on Dario's GTX460 or the 512 on your GTX580. I would imagine that would have a fairly noticeable impact. What I don't understand is the difference between your system and Dario's. I'm not sure if it comes down to differences in the FSX.CFG or what... Double check that each of the display settings tabs is identical to the screenshots in the guide. Also make sure you're both using 1680x1050 screen resolution on the graphics tab.I've been messing with the BP settings myself. I got about a 0.5fps increase by setting UsePools=0, but changing RejectThreshold had absolutely no effect whatsoever for me.

[dazz 69]

How did you go about disabling your Nvidia Inspector profile? I assumed removing FSX.exe from the MS Flight Simulator X profile would effectively disable it, so that's what I did.

Maybe that deletes FSX the profile but doesn't reset the driver? I really don't know. What's performance like with the profile enabled?This is what I did:-Exported my profile to a file FSX.nip (7th button from the left)-Restore current profile to NVIDIA defaults (2nd button with the Nvidia logo)

Also, should I have specified Anisotropic instead of Trilinear filtering in the benchmark???

I would think it won't make a difference, once the heavy stuff is offloaded from the GPU by disabling the Inpector profile, the test should be all about the CPU regardless. I could be wrong, but it's pretty easy to test anyway. I will do it now

[cmeeks 8]

Deleted

[alainneedle1 97]

OK I tried it, I understand i have to run 3 > 4 pass but I ran only one cause I have to go to bed, I'll try some more tomorrow., so far so good.No fsx.CFG tweak, fsx.CFG as it come, GPU to default, hardware in my sig.CPU @ 4677.4GHz, mem. at 2002MHz 7-7-7-21.

[SpiritFlyer 987]

It's graphics. I got a big perf boost disabling the Inspector profile and letting FSX handle FSAA & AFAre you guys sure you did that too?Just to clarify, I mean graphic settings in mi side. The test is perfectly set up for a CPU bound scenario

Well I thought I did it right! I guess I will have to eventually uninstall everything again and give it another go, but I think I will wait awhile to see how others with similiar machines do. With the Nvidia SLI tool and profiles it does not turn off and on the same way Nvidia Inspector does. In fact during the test the screen looked like it was still producing the same type of high quality image, so perhaps the C8xQ profile was still active. There certainly were no jaggies or obviously low anti-aliasing tell tales. We'll see how these numbers look when we get a lot more samples.There is always a reason for different numbers. Even still, it is hard to see how a 2600K at 5g with a 900Mhz 580 GTX can be outperformed by the same processor at 4.8g with a 460 GTX by a huge 20% margin, or by more than 50% over the exact same clocked chip with a 250 GS. Doesn't make any sense, not with those kind of numbers, but I suppose we will figure it out sooner or later.Kind regards,

[SpiritFlyer 987]

OK I tried it, I understand i have to run 3 > 4 pass but I ran only one cause I have to go to bed, I'll try some more tomorrow., so far so good.No fsx.CFG tweak, fsx.CFG as it come, GPU to default, hardware in my sig.CPU @ 4677.4GHz, mem. at 2002MHz 7-7-7-21.

Great numbers Alain. Let's see what the rest of the tests show. They should be excellent with your kind of horsepower.Kind regards,

[dazz 69]

Yeah Stephen, there's no way I can get higher FPS than you with the same CPU and a lesser video card. We'll figure it out for sure. I could even try a fresh FSX install if you guys want to try that at least to know what kind of figures we should be seeing. Then restore the back up and I'm up & running again. Your call guys