I wanted to put this info out there as hopefully being useful but also to pose a few questions i've run into. I've been working on a rebuild of my fsx rig from the ground up for Windows 8 x64 Pro RTM and i've done some benchmarking (yet again) and run into a few things, though mostly its going ok.

 

My rig consists of 3, 24" screens, for a max res of 5760x1200 with ati's eyefinity.. I run an ati 5850 1GB card and an i7 2600K with 16GB of memory (ddr3 1600). I've got a wide variety of addons i'm testing, but the base core consists of UTX, GEX 1.09, REX (4096), Orbx PNW, with some plane addons too. (I like to test in the C90B king, as it uses more fps than other VC's). I run in directx9 mode for now.. UT2 is installed but "disabled", though ut2services.exe is running in the background (this causes a crash when trying to use the fps_limiter.exe tool, which for now i cant run to test that mode).

 

The results of fsmark07 using the above (unless otherwise noted the benchmarks are using settings from the screenshots below) (target lock at 30fps):

 

I use fraps, set the benchmark to 420 seconds, Use the PNW addon as active (FTX central in North America mode)

 

Test1: 24.89 fps

CPU speed: 4.7ghz, Normal Autogen, Land Details Textures checked, LOD is set to 6.5

 

Test2: 23.47 fps

CPU speed: 4.7ghz, Dense autogen, LOD 6.5

 

Test3: 18.35 fps

CPU speed: 3.5ghz, Dense autogen, LOD 6.5

 

As you can see from the fps, going from nearly the 3.4ghz base i7 speed to 4.7 is only about a 5-6 fps increase (previous testing a year or more ago, i swear i got more like a 10 fps jump). Speedstep is on and c states are on as well (not sure if maybe a bios setting is different now than in past, or actually this increase of only 5-6 is typical i guess).

 

I found that the autogen slider left or right, each notch cost about 1 fps.

I also found that going DOWN to a 3xxx resolution didnt increase fps, it actually decreased my fps.. (which i found strange).

 

**I'm finding i have stutters a bit, at least with the fsxmark07 flight sequence, maybe its just the sequence.. as the flight starts to turn towards the city, my fps dips to around 16fps and there is a brief flash on the horizon in the clouds too.. i think maybe just something to do with the flight sequence itself.

 

**For reference, with these settings sitting on darrington runway, unpaused in the C90 king results in fps around 22-24 fps.. as go down the runway it dips to around 17.5 or so.. in air, generally 20-25 fps.. all the while locked at 30fps.

 

**If i try to increase the water to mid2x.. i lose an easy 3 fps on average as well. Shadows on ground forget it.. ill loose at least 7-9 fps on average.

**I'm using the shader mod 3 tweak too..

 

**Curious if anyone else has run this benchmark with similar hardware or could run it (420 seconds), matching settings to see if my numbers are consistent..

Overall windows 8 seems pretty solid with no major issues in general.

 

Here are my finalized settings i'm using and the fsx.cfg notes (these settings and tweaks are based on research i did on bufferpools and testing related to them all, this is where i ended up):

**I had to turn the AA to super-sample to get rid of shimmering buildings and trees.. (or turn off mipmap=6 completely).. I suppose I could lower the ati setting and leave mipmap off too, though there are still some shimmers and blinking textures.

 

 

Fsx.cfg is attached but some key settings i found to be the best:

 

 

TERRAIN_MAX_AUTOGEN_BUILDINGS_PER_CELL=1700

TERRAIN_MAX_AUTOGEN_TREES_PER_CELL=2000

 

ForceFullScreenVSync=1

 

bufferpools:

 

PoolSize=4000

RejectThreshold=524288

 

FIBER_FRAME_TIME_FRACTION=0.33

 

AffinityMask=14 (may be placebo effect)

poolsize and RejectThreshold can be there together.

AffinityMask=14 is ok for quad core, it's not placebo!

FIBER_FRAME_TIME_FRACTION=0.33 is default value, you get nothing with that value. Try 0.1, 0.2 etc.

You don't have highmemory fix?

 

 

Use this guide(no automatic tweaking tools), it's really the best one: http://forum.avsim.n...-guide-for-fsx/

 

You can backup your fsx.cfg, rebuild it and try this guide, if you dont like it, just copy and replace your old cfg

poolsize and RejectThreshold can be there together.

AffinityMask=14 is ok for quad core, it's not placebo!

FIBER_FRAME_TIME_FRACTION=0.33 is default value, you get nothing with that value. Try 0.1, 0.2 etc.

You don't have highmemory fix?

 

 

Use this guide(no automatic tweaking tools), it's really the best one: http://forum.avsim.n...-guide-for-fsx/

 

You can backup your fsx.cfg, rebuild it and try this guide, if you dont like it, just copy and replace your old cfg

 

Yeah I have both poolsize and reject together..

 

I modified my first post to include fsx config file (text version), but basically it contains a combination of settings i've read online in various guides, but i havent seen the link you posted here, so i'll check through there as well.

 

But overall, i'm just wondering if anyone else has run the fsxmark07 with pnw and gotten similar results (on similar hardware/settings).. and/or experiences in jumping up in the clock speed.. the margin of increase, 5, 10 fps etc..

 

Also, I'm not sure what the overall verdict is on using the fps limiter (external).. I cant even get it to run currently, as the ut2services.exe crashes when it launches fsx for some reason. Thats only a directx9 limiter, so once i tweak things per the dx10 guide, it will be useless. I read there is one other payware dx10 limiter out there, but not sure on how well it works as well.

 

The stutters i get with the fsxmark07 test locked at 30 go away if i set it to unlimited and things seem pretty smooth.

Yeah I have both poolsize and reject together..

 

I modified my first post to include fsx config file (text version), but basically it contains a combination of settings i've read online in various guides, but i havent seen the link you posted here, so i'll check through there as well.

 

But overall, i'm just wondering if anyone else has run the fsxmark07 with pnw and gotten similar results (on similar hardware/settings).. and/or experiences in jumping up in the clock speed.. the margin of increase, 5, 10 fps etc..

 

Also, I'm not sure what the overall verdict is on using the fps limiter (external).. I cant even get it to run currently, as the ut2services.exe crashes when it launches fsx for some reason. Thats only a directx9 limiter, so once i tweak things per the dx10 guide, it will be useless. I read there is one other payware dx10 limiter out there, but not sure on how well it works as well.

 

You can find a lot of fsx guides on internet, but really good guides that have made a fsx revolutions(each in its own time), are:

1. Nik's Guide http://www.simforums.com/forums/topic34141_post198187.html#198187

 

2. Bojote tweaking tool: http://www.venetubo.com/fsx.html

 

3. And now Word Not Allowed's Guide: http://forum.avsim.net/topic/370594-read-before-posting-software-hardware-guide-for-fsx/ which is the last one, and it's work's best for me and my friends who use fsx, so it's tested on different systems

 

I recommend you Word Not Allowed's guide.

 

But all this guides is for DX9 only, which almost all simers use becouse DX10 is not working well with fsx. If you want to try DX10, you can talk with Paul J there, he is the man: http://forum.avsim.net/topic/384221-a-how-to-go-dx10-doc/ and folow his guide.

 

About fps limiter, i can tell you only about DX9 mode, internal fps limiter is the best, you will find all info about this in Word Not Allowed's guide - there is the section about fps limiter and nvidia settings

As far as my current settings goes at 6.5 on LOD, i dont think i have many issues with blurries per the screenshot attached.

Edited September 23, 201213 yr by markm75

Keep in mind that even though it's "only" a 5fps difference, that's a 30% performance increase which is sizeable. At low framerates that makes all the difference.

I believe I made a breakthrough on determining the lesser fps in the benchmark...

 

I lowered ATI settings to 4x and 4x.. set AI to automatic, app controlled, lowered the supersample option down to the performance option (ill post these settings later).

 

Several tweaks I did NOT include in the fsx.cfg file either which i will also determine.

 

The biggest change was the Shader Mod3.. removing it.. not that it was a performance hit, but it caused a screen door effect/shimmering textures in many areas, which is why i had supersample cranked up (to remove the shimmer).

 

So without the shader mod I am getting around 29.21 fps on the 420 second test so far. I cant go to high 2x on water at all though.. this really drops my fps with or without the shader mod (so my setting above for water will remain, at least in dx9 mode).

Mark, how do you get that view that you have in the screenshot?

People expect too much from overclocking without looking at it logically.

 

You're overclocking by about 30%, and you're seeing a performance increase of nearly 30%.

 

Without doing tweaks, you're not likely to see any bigger improvement from overclocking alone.

Mark, how do you get that view that you have in the screenshot?

 

I just took a print screen, using fraps (set the print screen button to be that of print screen and jpg for output). The image had to be shrunk a bit for on here.. the original is 5760x1200.

 

I believe I made a breakthrough on determining the lesser fps in the benchmark...

 

I lowered ATI settings to 4x and 4x.. set AI to automatic, app controlled, lowered the supersample option down to the performance option (ill post these settings later).

 

Several tweaks I did NOT include in the fsx.cfg file either which i will also determine.

 

The biggest change was the Shader Mod3.. removing it.. not that it was a performance hit, but it caused a screen door effect/shimmering textures in many areas, which is why i had supersample cranked up (to remove the shimmer).

 

So without the shader mod I am getting around 29.21 fps on the 420 second test so far. I cant go to high 2x on water at all though.. this really drops my fps with or without the shader mod (so my setting above for water will remain, at least in dx9 mode).

 

I should mention that i did all this testing on a separate clean hard drive.. with just UTX, GEX, PNW and REX installed.

 

Changing the ati settings was the biggest factor.. I'll add the fsx entries i removed later (one was definitely not using bufferpools and also the affinity mask was not present, nor himemfix.. adding some of these back brought the avg fps down in combination with the ati settings below)..

 

**Right now i'm stuck on my original fully installed FSX hdd image.. it seems after i tried to uninstall the shader mod 3 hack something went amuck.. now I dont even get a virtual cockpit when i load up fsxmark07.. if i toggle fsx central back to Default i do however. I added the line for rebuilding the shader cache (and removed the one folder from the local appdata folder).. (after running the uninstaller for the shader mod).. so i'm not sure if i missed something or not. I'm fearing i may need to just do a complete rebuild on the fsx install which sucks.

 

Here are the settings in my ATI panel that did the trick on the "cleaner" hdd image (again i was able to verify using the recommended ati settings i had posted before and some of the common tweaks from online guides did bring my fps down a bit).. using my settings below i had hardly any stutter, motion was smooth, clear and great as well. Hence i'm back to what i believed my fps jump was before.. closer to 9-10 by going to 4.7ghz. And yeah, a 30% jump or more is a big difference in those instances where the system would normally take a hit, such as with rain or near a big city like seattle. I just knew from previous experience that I had a bigger jump somehow and have been trying to determine if it was win 8/atidriver related or fsx configuration based.

 

Sounds about right if you have fixed fps enabled. Let's say you'll get 1.4x more throughput with 4.7 from 3.4. about 10fps, but maybe about half of this will go into the look ahead buffer, so you see 5.

Just happy to report that since starting fresh (again).. i've got my fps at a decent rate.

 

With it set to unlimited and my sliders as noted before (except water at mid2x and terrain mesh at 1m not 5m, i also upped the AF to 8x from 4x to help with the runway looking blurry on some) i'm getting 32.5 fps on a 120 second benchmark using fsxmark07. If I switch to locked, I get 28.8 fps. No real shimmering (minor ones but you have to look hard to see).

 

If i induce heavy rain and lots of clouds and switch to the C90B king airplane and just fly around KSEA my fps will dip to around 17-19 worst case very close to the airport and then level out in the low to mid 20s with the rain still flowing (target 30).

 

Since KSEA is one of the worst, i'd say this is probably as good as its going to get in DX9 mode for my system. I found the right set of cfg tweaks that seemed to work best as well.

 

All of the above is on a clean install with only PNW, REX, UTX, Gex, Fsglobal 2010 installed.

 

I run at 5760x1200 now.. as noted before, I really didnt gain any fps by lowering down even into the 3xxxx x 1050 range for some reason.. I think after a certain point it makes no difference.

I run at 5760x1200 now.. as noted before, I really didnt gain any fps by lowering down even into the 3xxxx x 1050 range for some reason.. I think after a certain point it makes no difference.

 

Yeah, that's because your GPU has more than enough power for your FSX settings, therefore the bottleneck is your CPU/ram/everythingelse. FSX, being as old as it is, isn't that dependent on GPU power (until you start adding external tweaks like forced transparency AA etc)

Unfortunately the same 120 second test with my sliders and fsxmark07 at NIGHT were less than stellar.. where I got 28.4 in the day locked at 30, at night it was only 16.4 which is actually worse than in rain.

 

I believe there are topics out there about UTX being bad for fps at night however. What i've seen seems to indicate the only way around this is to turn off night lighting toggles in the UTX settings, but thats less than desirable, though as of now I do have both night lighting options enabled, I may try leaving just the main roads enabled and see how it fairs. I also have every single option enabled under every UTX menu too, which may increase the load on night lighting. I didnt select the day/night option in ftx however, perhaps with orbx there is no need for UTX (or utx lighting).

 

I took some tips from one guide online and turned off major road lighting.. in doing so my avg jumped from 16 to 21 fps on the fsxmark07 120 sec test. (t was around 18.4 fps with ftx set to night mode and major roads turned off in UTX.. visually there is a very minor difference with ftx day and night modes)

 

Here is a comparison of the 3 options:

 

#1: UTX with FTX to Night mode (UTX major roads off)

#2: UTX with FTX set to Day mode (Utx major roads off)

#3: UTX with FTX set to Day mode (major roads on)

 

Edited September 26, 201213 yr by firehawk44
Images removed. Exceed size 1600W max; Weight 400KB max

UTX Night Lighting is a massive FPS drain.