Sign in to follow this  
theskyisthelimit

My experiences with the EVGA 780Ti 3GB (vs 670FTW) and 5760x1200

Recommended Posts

I previously had an Nvidia 670 GTX FTW from EVGA (2GB)..

 

My system is as follows:

 

CPU: i7 2600k at 4.6 ghz (0.045+ offset)

RAM:  16 GB

OS:  Windows 8.1 x64 Pro 

HDD:  Sata III 64GB cache (for strictly the FSX portion) and a Sata III for the OS

Monitors:  24" Westinghouse L2410NM x 3 at 5760x1200 in nvidia surround mode

 

I used Word Not Allowed's AA Cloud test to compare performance on both the 670 and the 780ti with either 4xSupersampling or 4xSGSS (sparse grid super sampling).  All tests were in DX10 mode with the latest dx10 fixes.

 

Nvidia inspector is 1.9.72

Driver:  327.23 for the 670 tests and 331.65 for the 780Ti (had to install the latest for it to work).

 

For the OS for just the test, I had a clean OS, nothing installed.. just default weather and UTX and bufferpools=0 in both cases.

Frames are locked internally in FSX at 30 for all but one test.

 

Here is the relevant portion of my fsx.cfg for the test:

 

 

[DISPLAY.Device.NVIDIA GeForce GTX 780 Ti.0.0]
Mode=5760x1200x32
Anisotropic=1
AntiAlias=1
 
[bUFFERPOOLS]
UsePools=0
 
[GRAPHICS]
AC_SELF_SHADOW=0
AIRCRAFT_REFLECTIONS=1
AIRCRAFT_SHADOWS=1
ALLOW_SHADER_30=1
COCKPIT_HIGH_LOD=1
D3D10=1
EFFECTS_QUALITY=2
ForceWindowedVSync=0
GROUND_SHADOWS=0
HIGHMEMFIX=1 
IMAGE_QUALITY=0
LANDING_LIGHTS=1
 
 
MultiSamplesPerPixel=4
MultiSampleQuality=0
 
NUM_LIGHTS=8
See_Self=1
TEXTURE_MAX_LOAD=2048
TEXTURE_QUALITY=3
SHADER_CACHE_PRIMED_10=1693500672
SHADER_CACHE_PRIMED=1693500672
 
[Display]
BLOOM_EFFECTS=1
SKINNED_ANIMATIONS=1
TEXTURE_BANDWIDTH_MULT=120
TextureMaxLoad=9
TransitionTime=4.000000
UPPER_FRAMERATE_LIMIT=30
WideViewAspect=True
 
[TERRAIN]
AUTOGEN_DENSITY=3
BATCH_AUTOGEN2=1
DETAIL_TEXTURE=1
LOD_RADIUS=7.500000
MESH_COMPLEXITY=90
MESH_RESOLUTION=24
//SWAP_WAIT_TIMEOUT=2  
//TERRAIN_MAX_AUTOGEN_BUILDINGS_PER_CELL=3000     //897
//TERRAIN_MAX_AUTOGEN_TREES_PER_CELL=4500  //1100
TEXTURE_RESOLUTION=29
WATER_EFFECTS=4
 
 
[sCENERY]
//ObjectsToBatchPerFrame=50 //new
DAWN_DUSK_SMOOTHING=1
IMAGE_COMPLEXITY=4
LENSFLARE=1
//MAX_ASYNC_BATCHING_JOBS=3
//SmallPartRejectRadius=1  //4 b4
 
[Main]
DisablePreload=1
FIBER_FRAME_TIME_FRACTION=0.15 //0.05  
 
 
 
[TrafficManager]
AIRPORT_SCENERY_DENSITY=1
AirlineDensity=0
FreewayDensity=7
GADensity=0
IFROnly=0
LeisureBoatsDensity=0
ShipsAndFerriesDensity=0
 
[Weather]
CLOUD_COVERAGE_DENSITY=8
CLOUD_DRAW_DISTANCE=3
DETAILED_CLOUDS=1
DisableTurbulence=0
DownloadWindsAloft=0
DynamicWeather=0
THERMAL_VISUALS=0
WindshieldPrecipitationEffects=1
 
 
And here is my nvidia settings in inspector for 4xsupersample (switch to 4xSGSS for the other test)
nvidia_20131114_211656.png
 
With the above settings I got the following results with the AA cloud test by Word Not Allowed (see his wordpress site for the download if you want to try it yourself).. I used fraps set the timer at 60 seconds to test.. again FSX locked at 30 fps
 
670GTX FTW 2GB:
 

8.5 FPS (4xSGSS)

20.8 FPS (4xSupersampling)

 

Then for the big daddy card.. the 780Ti..

11.03 FPS (4xSGSS)

29.26 FPS (4xSupersampling)  (if i unlock frames it goes to 29.6 fps)

 

Now for general use in FSX on my normal install I found the frames to be better overall.. for instance.. the Miami 2012 addon at night was previously coming in around 15 fps with the 780 and 4xsupersample it was clocking in around 22 fps.. meanwhile KSEA with heavy weather in one instance with just a default cessna used to come in around 14 fps.. it now holds 26 fps.. so it is better in this sense.. but at a cost of living with minor flickering with 4xsupersampling.  4xSGSS is still a problem..

 

Conclusion..

 

This is my take at this point, though maybe more tweaking is needed on the driver side..  At a price point of $739 compared to what is now around $300 for the 670 FTW, i'm not sure if its worth the price difference, though as noted with my general FSX use, perhaps, if your willing to accept living with minor flickering and 4xSupersampling..

 

Again I have more testing and tweaking to try, but unless someone has a miracle boost cure here, I dont expect these numbers to climb much higher, though ill update here if i make more progress.

 

Looking backward though and comparing to Kostas result in comparing a 580, 680 and Titan, it still seems like the 780ti should have come in somewhere between the 680 and Titan..

 

He got the following:

 

10 fps on 580

14 fps on 680

20 fps on the Titan.

 

I would have expected values in the range of 16-18, though the test parameters may have been slightly different in his test (dx10 or dx9?)

  • Upvote 1

Share this post


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

And people still say FSX is the future... If a card like this can't handle more than 11FPS, no card ever will!!! This thing is a beast.

 

Let's wait for the other about to come DX11 Simulator, I hope it handles these newer cards better!

 

Edit: Didn't realize you run at absurdly high resolutions!! Maybe that's why it's so low.

  • Upvote 1

Share this post


Link to post
Share on other sites

And people still say FSX is the future... If a card like this can't handle more than 11FPS, no card ever will!!! This thing is a beast.

 

Let's wait for the other about to come DX11 Simulator, I hope it handles these newer cards better!

 

Edit: Didn't realize you run at absurdly high resolutions!! Maybe that's why it's so low.

 

Yeah my three 24's force a native of 5760x1200.. so its up there.  Even with the 670 2GB if you lower the res to say 3840, it worked much better, but due to native resolution issues, it looks like crap.. alternative, downsize 3 screens and get new ones, go to one screen.. cost for cost, probably about the same differential by just getting this card (for me).

 

That 10 fps is a beast of a test though.  These cards do help a little, but yeah still CPU locked issues drive most of the performance even at 4.6ghz.

 

Actual usage is 22-29 fps (locked at 30) with the 780ti with 7% road traffic and minimum on static airplanes at airports.  I was 14-30 with the 670 2GB at this resolution.

 

So going from say 14 fps worst, to maybe 21 fps worst (night mode) is pretty good, but by no means worth the price hike.

 

This card supports directx 11.2 though.. it will be interesting to see how P3D v2.0 stacks up.

 

I'm going to test a 780 FTW and see how its numbers stack up as well.. statistical fun :)

Share this post


Link to post
Share on other sites

Yikes this test result is kinda a bummer for me - we have about the same setup (processor, clock, screens,  etc.)  and I just ordered the 780 non ti.

 

Found a Titan the next day used for the same as I paid for the 780.

 

Maybe a second card (in my future) might help...?

Share this post


Link to post
Share on other sites

 

Yikes this test result is kinda a bummer for me - we have about the same setup (processor, clock, screens,  etc.)  and I just ordered the 780 non ti.

 

Found a Titan the next day used for the same as I paid for the 780.

 

Maybe a second card (in my future) might help...?

 

I've read that say two 670 4GB can be a real benefit..

 

Are you using the titan right now? Would be curious at our high res, what fps you get on the Word Not Allowed AA cloud test with 4xSGSS.  Thats a steal if you managed that.. the 780 FTW is around $540 shipped.. I've seen titans averaging sold, used on ebay for around $750 (average).

 

I will be testing the 780 FTW over the weekend to see how it stacks up.  Eventually i'll put all this into one article on my website with a chart/table with pics to better see how all 3 cards stacked up.

 

The test clouds flight situation is here if anyone wants to give it a shot on their machine (using Fraps and a 60 second benchmark). And if possible try it using my settings in the first post (sliders and nvidia NI).

Share this post


Link to post
Share on other sites

Running 4xSGSS with Word Not Allowed's cloud test is just suicidal anyways. It should melt any card at that resolution.

Share this post


Link to post
Share on other sites

Running 4xSGSS with Word Not Allowed's cloud test is just suicidal anyways. It should melt any card at that resolution.

True. But my average of 22-29 will certainly drop by a large margin with general use and 4xSGSS even on the ti. Hence I stick with 4xsupersample. I'll also try 2xSGSS and see how that performs.

 

 

Share this post


Link to post
Share on other sites

2xSGSS should be ok. How can you stand the the shimmering with 4xSS?

With my system I barely see any significant shimmering at 4ss(for some reason). Part of a building may flicker and static airplanes shimmer way off in distance. Trees are not that bad. For the performance drop in moderate conditions to me it was better to use 4ss. Plus going from 20 to nearly 30 on the AA test at that level is around a 30% jump. Not bad.

Share this post


Link to post
Share on other sites

I do not have the Titan - the guy contacted me back one day after I ordered the 780 Classified. It was on Craigslist (order at your own risk type thing). It is gone now.

 

In all honesty I will probably not go for the leading edge of settings. I will just setup to a point and leave it alone. Last time I was into FSX I tweaked all of the time and it became a habit I really do not want to get back into.

 

I see you have all of the Saitek panels - I have them all too and just made a simple pit for them. I was tired of setting them up and taking them down so made them into a cabinet. When I get set back up (end of next week) I will post some pictures. The whole thing cost less than $100 (not including panels) and is 100 times more fun now.

Share this post


Link to post
Share on other sites

Well your lucky, anything less is unacceptable for me. Maybe it's different with multi monitor setups and higher resolutions. I only run one monitor at 1920

Share this post


Link to post
Share on other sites

I do not have the Titan - the guy contacted me back one day after I ordered the 780 Classified. It was on Craigslist (order at your own risk type thing). It is gone now.

 

In all honesty I will probably not go for the leading edge of settings. I will just setup to a point and leave it alone. Last time I was into FSX I tweaked all of the time and it became a habit I really do not want to get back into.

 

I see you have all of the Saitek panels - I have them all too and just made a simple pit for them. I was tired of setting them up and taking them down so made them into a cabinet. When I get set back up (end of next week) I will post some pictures. The whole thing cost less than $100 (not including panels) and is 100 times more fun now.

Ah nice. Yea I'd love to see pics. I love the panels. Mine stay up under my screens 24-7. Btw I use SPAD for the driver and it fixed some issues. On the tweaking I agree. I find myself mostly tweaking and blogging or forum chatting about it. I guess it's a hobby now. However it felt spending more on the TI if I could get past the tweaking is worth it. Time is money thing :).

Well your lucky, anything less is unacceptable for me. Maybe it's different with multi monitor setups and higher resolutions. I only run one monitor at 1920

Yep I bet is it the higher resolution. Saves slightly on the need for higher AA.

Share this post


Link to post
Share on other sites

FSX just can't take advantage of modern GPUs. You can buy a $700 GPU that shows amazing performance in any other game yet just a 5-10 FPS improvement in FSX. That's why I think buying high-end GPUs for FSX is useless. The cost just doesn't justify the performance advantage.

Share this post


Link to post
Share on other sites

FSX just can't take advantage of modern GPUs. You can buy a $700 GPU that shows amazing performance in any other game yet just a 5-10 FPS improvement in FSX. That's why I think buying high-end GPUs for FSX is useless. The cost just doesn't justify the performance advantage.

 

I couldn't agree more. I would never select hardware of any sort for just FSX. I choose for many other software reasons. FSX maybe gets 5-10% in the consideration process.

Share this post


Link to post
Share on other sites

I couldn't agree more. I would never select hardware of any sort for just FSX. I choose for many other software reasons. FSX maybe gets 5-10% in the consideration process.

 

FSX has a very finicky taste for hardware. I'm sure Aces built it with 8GHz single-core CPUs in mind, but instead dual-cores happened. SP1 adds some multithreading, but the load on the first thread is extremely heavy.

 

I'm placing high hopes on Prepar3D 2.0. With proper multithreading and GPU usage, the performance might be as high as Microsoft Flight.

Share this post


Link to post
Share on other sites

Doing some more generic flying with the 780ti and still in 4xSupersample mode..

 

The 737-800 NGX at Toulouse addon airport in the dark is clocking in around 22 fps.. switch to daytime with some clouds (with breaks).. getting around 17-22 as well (int frame lock here).. I havent run this same area under the 670 FTW 2GB to compare (as of yet)

 

EDIT: restarted the sim, this time the numbers went up.. now at least in the day test i'm 22-29 fps.. very acceptable.. occasional micro studders with locked internal mode, nothing too bad.. if i use external lock the numbers go up slightly.

 

GPU load hit 99% while memory load is at 2321MB out of the 3GB (bufferpools=0).. GPU temp hit 79C with a fan speed of 74%. According to what I read the throttle point on the 780ti is 83C while the Titan and 780 is 80C. Most benchmark tests done by independent reviews seem to show a peak of 82C

Share this post


Link to post
Share on other sites

Doing some more generic flying with the 780ti and still in 4xSupersample mode..

 

The 737-800 NGX at Toulouse addon airport in the dark is clocking in around 22 fps.. switch to daytime with some clouds (with breaks).. getting around 17-22 as well (int frame lock here).. I havent run this same area under the 670 FTW 2GB to compare (as of yet)

 

EDIT: restarted the sim, this time the numbers went up.. now at least in the day test i'm 22-29 fps.. very acceptable.. occasional micro studders with locked internal mode, nothing too bad.. if i use external lock the numbers go up slightly.

 

GPU load hit 99% while memory load is at 2321MB out of the 3GB (bufferpools=0).. GPU temp hit 79C with a fan speed of 74%. According to what I read the throttle point on the 780ti is 83C while the Titan and 780 is 80C. Most benchmark tests done by independent reviews seem to show a peak of 82C

EDIT: restarted the sim, this time the numbers went up.. now at least in the day test i'm 22-29 fps.. very acceptable.. occasional micro studders with locked internal mode, nothing too bad.. if i use external lock the numbers go up slightly.

Thanks for the update - Are you still running - FTX, GEX, FS Global 2010, UTX?

 

Yep.. see just now post.. the fps jumped up.. not sure what i did besides restart things.. looking good with this test anyway :)

Share this post


Link to post
Share on other sites

Some updated news on my testing..

 

I opted to try the same tests on Windows 7.. just to see if there was any change.. i found an increase of about 1.2 fps in Windows 7, but that could have been due to a number of factors.. plus my 7 box is a full blown image of FSX.. so i'd call it a wash.

 

I ran more tests.. this time comparing all 3 cards (780 FTW as well).. summary again.. my 670FTW is basically almost the same as as a 680 at least in PCB terms and overclocking by the factory..  I tried internal vs external locking.. for me it seemed like external had less micro pauses and about 0.5 fps faster on average.

 

Bare Win 8.1 minimal FSX install, int locked at 30/DX10-5760x1200; LOD 7.5; 2048 (see full config in first post)

 

Test Cloud AA

4xSGSS 670FTW:   8.56 

4xSGSS 780FTW:   9.21

4xSGSS 780TiSC:  11.03  (10.75 with rex and full blown win 8 image of fsx with several addons, 10.78 if set to unlim frames)

 

Test Cloud AA

4xSS  670FTW:  20.80 

4xSS 780FTW:   26.61

4xSS 780TiSC:        29.26   (28.70 with rex and full blown win 8 image of fsx with several addons, 29.11 with NI external lock at 30)

 

Performance Drain 

4xSS 670FTW:   21.67  (full win 8, fsx install with rex etc)

4xSS 780FTW:   NA (didnt run yet)

4xSS 780TiSC:        NA (didnt run yet)

 

For another test i tested the NYCX addon, which i cant really get above 13 fps no matter WHICH card i use, when close to the city (and no matter how much i tweak the complexity/ag).. i set throttle to full on the tecnam p92 eaglet and did a 25 second timing test when approaching the downtown zoo area from a short distance away (high scattered clouds):

4xSS 670FTW:  12.52 

4xSS 780FTW:   NA (didnt test)

4xSS 780tiSC:    12.60

 

As you can see even with a $400 more video card such as the 780ti.. it cant add much of a benefit in this situation at least..  There should be some situations where having that horsepower can help, but its my opinion with FSX being so CPU bound, 90% of the time there may be no benefit.. though i have some scenarios to test.

 

A bit of a bummer.. hoping that the 780ti would churn out 5760x1200 a bit better in all cases, but so far not the case.. the AA cloud test does show a 10 fps gain, but thats a more rare instance I believe.  I tried some other scenarios.. like the previously mentioned France test with the NGX.. couldn't see a huge difference between the 670 and 780ti.. maybe 4-5 fps if that.

 

I'm going to try some more scenarios to see for sure.. but so far.. it seems if your card is a 670/680.. its pretty beefy as is.. especially if you use lower resolutions.. my needs are for high resolution, since i have the 24's.. so that makes it harder and more demanding and shows less of a benefit with the 780ti.. lower that resolution its a whole other story.

 

The 780ti could prove useful with Prepar3D v2.0 down the road, so it may be worth holding onto.

 

Worth noting but I do run the EVGA Precision software during testing.. trying to keep the 670 and 780 at 70C or less to prevent throttling and the 780ti at 83C or less all using the fan/temperature curve settings.

 

Here were my 3dmark11 Performance scores for each (at 4.6ghz i7 2600k), the 780ti wasn't immensely higher.. i've seen some web reviews with numbers around 13,500, mine didnt get that high even with my OC'd cpu for some reason (update:  my board is the p8p67 which is pcie 2.0, the card is rated at 3.0 which may or may not bring the numbers down a tad, the 780ti fill rate is 210 GT/sec so even with 3.0 only a 4% improvement in 3dmark11):

 

670FTW: P8685

780FTW: P12074

780Ti:      P12813

 

It could prove an interesting experiment to get another 2GB 670FTW and see how it performs on these same tests in SLI mode with my resolution.  Some SLI scores i've seen on the web are in the 14,680 realm with SLI, which would be higher at least 3dmark11 wise than the 780ti.

  • Upvote 1

Share this post


Link to post
Share on other sites

Thanks for the updates buddy!

 

My new 780 Classified just arrived with three new 27" monitors - Ahhh the office smells of new computer equipment - love it!

 

Not hooked anything up yet going to wait until I turn my current work project over late tomorrow before messing around in the box. Priorities ya know.

 

Sounds like the new Asus Black is at Newegg and shipping. I plan to give it a month (na I will fold in a couple of weeks) then time to update my processor (i7-4930K) and MB.

 

I am getting the new gear mainly for work but it will be interesting to see what all of the new hardware does in FSX.

 

Going to OC the 780 in a couple of weeks once I am sure the GPU is good. Nice to be back to Nvidia again.

 

Again thanks for the testing updates reps added. Makes for a great read while I waited.

Share this post


Link to post
Share on other sites

 

 


Worth noting but I do run the EVGA Precision software during testing.. trying to keep the 670 and 780 at 70C or less to prevent throttling and the 780ti at 83C or less all using the fan/temperature curve settings.

 

Don't hold the 780 back to a lower temperature like 70C, by doing that you're likely limiting the performance it's going to give you in the sim and not letting it stretch it's legs so to speak.

 

If you're letting the 780ti max out at it's throttle point of 83C, then let the 670 and 780 go to their throttle points too so you can get a more fair and accurate representation of the results. Otherwise your results may be skewed.

 

Hell, you can let the 780 run up to 90-95C if you're keen on overclocking. Not suggesting you do this though. lol. If I'm overclocking my 780 I set my temp target at about 84C and have no throttling whatsoever.

 

Just my 2c.

 

Also, don't forget about fan profiles. These can impact things too. If you make a change on one card, make the same change on the other.

  • Upvote 1

Share this post


Link to post
Share on other sites

Thanks for the updates buddy!

 

My new 780 Classified just arrived with three new 27" monitors - Ahhh the office smells of new computer equipment - love it!

 

Not hooked anything up yet going to wait until I turn my current work project over late tomorrow before messing around in the box. Priorities ya know.

 

Sounds like the new Asus Black is at Newegg and shipping. I plan to give it a month (na I will fold in a couple of weeks) then time to update my processor (i7-4930K) and MB.

 

I am getting the new gear mainly for work but it will be interesting to see what all of the new hardware does in FSX.

 

Going to OC the 780 in a couple of weeks once I am sure the GPU is good. Nice to be back to Nvidia again.

 

Again thanks for the testing updates reps added. Makes for a great read while I waited.

 

Wow, that sounds like a fantastic build, with 27" screens x 3 to boot.. amazing!

 

I'd be real curious to see how your 2600k at 4.6 going to the 6 core works out.. technically if you overclock that 6 core its Passmark CPUMarks should surpass the i7-2600k easily.. however i wonder if that would really matter for FSX.. since it just likes raw power to power the fsx.exe process.. all things held equal (same video card) i'd be real curious.. because i was considering switching to ivy bridge or haswell and going to a better proc one day soon.

 

Yeah I think your 780 Classified will do fine.

 

Happy building! :)

Don't hold the 780 back to a lower temperature like 70C, by doing that you're likely limiting the performance it's going to give you in the sim and not letting it stretch it's legs so to speak.

 

If you're letting the 780ti max out at it's throttle point of 83C, then let the 670 and 780 go to their throttle points too so you can get a more fair and accurate representation of the results. Otherwise your results may be skewed.

 

Hell, you can let the 780 run up to 90-95C if you're keen on overclocking. Not suggesting you do this though. lol. If I'm overclocking my 780 I set my temp target at about 84C and have no throttling whatsoever.

 

Just my 2c.

 

Also, don't forget about fan profiles. These can impact things too. If you make a change on one card, make the same change on the other.

 

Gave it a shot and in both cases my numbers were about the same.. still.. point taken :)

 

I've tried OC'n video cards and even the 670FTW before.. never saw a real change in fps for fsx anyway.

Share this post


Link to post
Share on other sites

 

 


Gave it a shot and in both cases my numbers were about the same.. still.. point taken :)



I've tried OC'n video cards and even the 670FTW before.. never saw a real change in fps for fsx anyway.

 

I just recently tried overclocking my 780 and in testing the average FPS in FSX increased by about 2FPS. So whilst not mind blowing it can make a difference, and the little 1-2FPS from here, there and everywhere all add up.

 

This was tested using the FSXMark11 test flight plan with the average calculated based on the last 4 flights out of 5 flights conducted (the first flight you load in FSX will always have lower FPS due to tecture loading etc, so this result gets discarded)

Share this post


Link to post
Share on other sites

Yes it will be interesting build. I mainly want to watch the six processors chew through rending on 3d, video, and geo-data processing.

 

I have to say the 27s are a bit of overkill now that I have them all here. It is a super wide span. They span over 6 feet. So much so that maybe TrackIr would not even be necessary.  I kinda like overkill in everything I do.

 

Anthony that is interesting and that you pump out a couple of extra fps from adjust a few numbers in the setting- that is well worth it. I see you have 2 780s I hope to do that too after the new year. Upgrading the MB to a 3.0 PCI I hope will help bottlenecks.

 

Looks like Newegg is out of stock now on the Blacks.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this