Back to FSX and new to DX10 - perfs...

[cvearl 60]

A tip when benchmarking.... The mouse cursor showing on screen in FSX eats 3 to 5 fps. Known fsx issue. Buddy with the GTX680 taught me that. Another tip. Of the 3 people I know with the same CPU as you, all run 4.4Ghz. They simply bumped the multiplier to 44 on the motherboard (all cores) and disabled that CPU setting so that they are at 4.4 pretty much 24/7.

 

That's some heavy scenario with 75% traffic. Your last two situations are certainly CPU bound. We have nearly identical systems. I think you can get better performance though. I would like to compare. But we need similar scenarios. I don't have NGX but I have the JS41.

 

1. That LFPG. Is that payware?

 

2. What is your cloud density and draw distance set to in weather tab of fsx.

 

3. Is Hyperthreading on or off in BIOS?

 

4. Besides 75% traffic, what's ground vehicles and road traffic set to? They are major fps eaters.

 

C.

[Osirith 14]

A tip when benchmarking.... The mouse cursor showing on screen in FSX eats 3 to 5 fps. Known fsx issue. Buddy with the GTX680 taught me that. Another tip. Of the 3 people I know with the same CPU as you, all run 4.4Ghz. They simply bumped the multiplier to 44 on the motherboard (all cores) and disabled that CPU setting so that they are at 4.4 pretty much 24/7.

 

That's some heavy scenario with 75% traffic. Your last two situations are certainly CPU bound. We have nearly identical systems. I think you can get better performance though. I would like to compare. But we need similar scenarios. I don't have NGX but I have the JS41.

 

1. That LFPG. Is that payware?

 

2. What is your cloud density and draw distance set to in weather tab of fsx.

 

3. Is Hyperthreading on or off in BIOS?

 

4. Besides 75% traffic, what's ground vehicles and road traffic set to? They are major fps eaters.

 

C.

Hi,

 

thanks for helping :rolleyes:

Mouse is maybe seen because I made a screenshot with a specific capture tool. Should not be present before pressing PrtScrn key, so no effect on FPS.

 

About the CPU and overclocking : need to try to o/c a bit more, will do it as soon as I find some spare time ...

1. LFPG is Aerosoft Megaairport LFPG. Payware. I can switch to a standard FSX airport or anything freeware to help compare with your own settings.

 

2. Cloud Density should be maximum. Cloud Distance : need to check at home, will need to wait this evening (CET-West European time zone).

 

3. Hyperthreading should be off, hence the Affinity to 14. Did not see a lot of differences with Hyperthreading On/Affinity 254.

[Osirith 14]

4. Besides 75% traffic, what's ground vehicles and road traffic set to? They are major fps eaters.

 

Not sure while not in front of the PC, but Airport Vehicle Density should be set at Low or something (below Medium) and Road Véhicles should be between 10 and 20%.

 

Also, Air traffic is set at 0% inside FSX, and 75% inside UT2.

[cvearl 60]

Not sure while not in front of the PC, but Airport Vehicle Density should be set at Low or something (below Medium) and Road Véhicles should be between 10 and 20%.

 

Also, Air traffic is set at 0% inside FSX, and 75% inside UT2.

I'm guessing to simulate the number of planes on the ground you showed there, I would need to set FSX to 100% ? I have never seen that many planes in one place in the sim myself.

 

Hey. Do you own any ORBX scenery? Recent HD Carenado planes like B200? How about RealAir Duke Turbine?

 

C.

[Osirith 14]

I'm guessing to simulate the number of planes on the ground you showed there, I would need to set FSX to 100% ? I have never seen that many planes in one place in the sim myself.

 

Hey. Do you own any ORBX scenery? Recent HD Carenado planes like B200? How about RealAir Duke Turbine?

 

C.

 

Not sure about traffic setting inside FSX. I guess 100% should be close ? Not sure either is UT2 is more resource-hungry for managing traffic (for example, it uses actual jet routes, no straight lines from airport A to airport B)

 

And no. I do not own any ORBX scenery (except a few freeware samples), not Carenado planes, nor RealAir Duke... So far, for FSX I only bought PMDG 737NGX and VRS F-18E airplanes, and next on the list will be PMDG 777 (cannnnnn't wait, seeing today screnshots).

[Osirith 14]

Update : Airport Vehicle Density should is set at Low and Road Véhicles is set at  20%.
 

Cloud Distance is set at minimum (60 nm).

[cvearl 60]

Update : Airport Vehicle Density should is set at Low and Road Véhicles is set at  20%.

 

Cloud Distance is set at minimum (60 nm).

 

For a common test, we can do the FSXMARK11 thing Word Not Allowed (at his blog about the Titan) tested the GTX580, 680 and Titan. Do you have FRAPS? It's free. Well... Just read the included PDF for instructions. Good test. It was one of the ones he used.

 

FSXMARK11 - http://forum.avsim.net/topic/328262-fsxmark11/

The download - https://dl.dropbox.com/s/z898vaszoh2vyqk/FSXMark11.zip?dl=1

 

NOTE: Do NOT reinstall FSX like they are asking. That was for a project they were doing (clean install result only). Just match Kostas FSX MENU settings (not FSX.CFG tweaks) and lets compare there. Here are his results and his settings. His tests were all in DX9 mode BTW. I will provide DX10 and DX9 results (using Bojote's DX9 SM 3.0 clouds rendering path aka ATI mode clouds). As for his result, here is a rough translation from nVidia AA modes to ATI. For the test, lets just focus on 4xSSAA.

 

2xSGSS = 4xSSAA (based on the premise he is using 8xS for AA + 2xSGSS transparency) 4xSSAA is just about the same.

8xSGSS = 8xSSAA (ATI 8xSSAA is nearly indistingishable in side by side's versus either 4xSGSS adn 8xSGSS). too high for most systems.

 

 

[cvearl 60]

Video Card - Gigabyte HD 7970 3GB OC (Ghz clocked)

 

Ok so I set up exactly as Word Not Allowed did in his GTX680/Titan tests See screenshots in post just before this. Here are HD 7970 3GB results. LOD Radius 6.5, 1920x1080x32 res.

 

DX9+SM 3.0 Clouds and DX10 Preview

 

DX9 FSMark11 4xSSAA = 36 fps

DX9 FSMark11 8xSSAA = 25 fps

DX10 FSMark11 4xSSAA = 45 fps

DX9 Test Clouds AA 4xSSAA = 32 fps

DX9 Test Clouds AA 8xSSAA = 18 fps

DX10 Test Clouds AA 4xSSAA = 25 fps

DX9 Overcast Performance Drain Test 4xSSAA = 36

DX9 Overcast Performance Drain Test 8xSSAA = 22

DX10 Overcast Performance Drain Test 4xSSAA = 38

 

Compare...

 

[MichaelMoe 1]

So you point and goal in this performance tests is to clarify that getting a new up to date ATI or GTX Card will improve the overall FPS in FSX ?

 

I am thinking of upgrading soon to this HD7970 3GB from a ATI58701GB mainly because of FSDT pure DX10 code scenerys that eats a lot of VideoRam in HD textures.

 

Hopefully i will see a small FPS gain elsewhere 

 

Michael Moe

[cvearl 60]

So you point and goal in this performance tests is to clarify that getting a new up to date ATI or GTX Card will improve the overall FPS in FSX ?

 

I am thinking of upgrading soon to this HD7970 3GB from a ATI58701GB mainly because of FSDT pure DX10 code scenerys that eats a lot of VideoRam in HD textures.

 

Hopefully i will see a small FPS gain elsewhere

 

Michael Moe

Ya more or less. I had a 660ti and when things got heavy (payware airport, HD plane, bad weather), fps STILL tanked into the high teens in some cases. 7950 helped, 7970 helped. I can't afford the Titan. ATI's next card is not for another 6 months at least.

 

C.

Two more pics showing extreme settings and conditions, and low frame rates.

However, this is a quite exceptional case and I could live with that. Most of the time, I will only experience frame rates around 15 during taxy and very "rich/complex" areas; so it is manageable.

The first pic showing 20FPS is almost misleading as frame rate in this view can drop as low as 12 for a few seconds.LFPG, External view, PMDG737, ugly weather, 75% traffic, in flightLFPG, VC view with HGS on, PMDG737, ugly weather, 75% traffic, in flight

Yup. That's extreme. I think your squeezing all you can out of ANY system. I think you need to bump that CPU to 4.4. In your BIOS, from 40 to 44. Pretty sure you'll only need a multiplier bump. Doubtful it will need more voltage or anything like that.

 

C.

[Osirith 14]

Hi Charles,

 

not enough time to did benches, however I pushed the CPU to 4.6 Ghz and it's seems totally stable.

 

Tried Prime 95 for 15 mns, max temp reach was 70°C (158 F) which seems not too bad.

 

Aftwerwards, did a complete flight from KLAX to KSFO, starting with Cold & Dark cockpit, and temperature never reached 70 (There is a visual and sound warning at this temp). Frame rate was very good, never below 20 on ground with 75% UT2 traffic and some specific scenery.

[cvearl 60]

Hi Charles,

 

not enough time to did benches, however I pushed the CPU to 4.6 Ghz and it's seems totally stable.

 

Tried Prime 95 for 15 mns, max temp reach was 70°C (158 F) which seems not too bad.

 

Aftwerwards, did a complete flight from KLAX to KSFO, starting with Cold & Dark cockpit, and temperature never reached 70 (There is a visual and sound warning at this temp). Frame rate was very good, never below 20 on ground with 75% UT2 traffic and some specific scenery.

 

That's pretty impressive for the type of flying you are doing with all that crazy traffic running around.

 

I have come to understand that that FSXMark11 benchmark is mainly CPU and MEMORY speed bound. Faster cards have little effect. Still an interresting test though. But ya. It's all about CPU in that test. You would probably get a good result with that CPU at 4.6 fo sho.

 

C.

[Osirith 14]

I still need to redo a test at LFPG. I suspect that UT2 traffic is much heavier there and Aerosoft scenery takes a heavier toll on frame rates. Also, my californian flight today was with good weather... So I am not sure to have grabbed a large amount of frames with the new overclock setting. Anyway, going from 4.0 to 4.6 Ghz should in theory at least gives 15% more fps. Which means 1 to 3 more frames...