Jump to content
iMolestCats

CM performance thread

Recommended Posts

Thanks Phil. For what it's worth I dropped the settings in game for models and textures down to 'Fastest' and got an avg of 44 FPS. I assume that backs up the OpenGL theory?

Also that's interesting about Shogun 2, never realised DirectX and OpenGL would cause such a difference. But I'm not a programmer just a happy consumer who likes sending pixels to their deaths. :D

Share this post


Link to post
Share on other sites
...Upgrading may not significantly increase your frame rates. What it MAY do is make you more able to play larger scenarios more smoothly. That's not as counter-intuitive as it sounds. Averaging even 1-2 more FPS more will help a lot if there's a heap of processing happening and everything takes a dip.

Understood and would be happy to take part in a similar test targeting a large/huge scenario for further comparison purposes.

Share this post


Link to post
Share on other sites
Guest
Thanks Phil. For what it's worth I dropped the settings in game for models and textures down to 'Fastest' and got an avg of 44 FPS. I assume that backs up the OpenGL theory?

Going down to "Fastest" for models cuts rendering load down incredibly, and also lowers CPU load a bit. It smooths things out quite a bit - the CPU and GPU aren't both doing huge amounts of work at once. In terms of the OpenGL theory - on Fastest I think you should probably be seeing higher frame rates. It would be interesting to see what frame rates other folks got on Fastest.

With a card like yours you shouldn't see much of a hit from texture quality being at Better or Best. Most modern cards can handle the higher texture quality settings.

Also that's interesting about Shogun 2, never realised DirectX and OpenGL would cause such a difference. But I'm not a programmer just a happy consumer who likes sending pixels to their deaths. :D

Me too, when I'm not programming! :)

DirectX vs OpenGL is a huge difference in terms of driver quality. You'll definitely see more bugs in OpenGL.

In terms of Shogun vs. CM, though, it's largely down to how the games work - Shogun can do many more optimizations than we can due to how they render their troops and battlefields, and their camera setup, and they also do much less pathfinding, rendering, and AI work. A lot of the character of Combat Mission relies on us doing that extra work, though - all that AI and pathfinding is what makes CM the tactical game that it is.

Understood and would be happy to take part in a similar test targeting a large/huge scenario for further comparison purposes.

That would be cool. Maybe a follow-up to this thread?

Share this post


Link to post
Share on other sites
Understood and would be happy to take part in a similar test targeting a large/huge scenario for further comparison purposes.

I volunteer. Anyone got a half finished game of 'Colossal Crack' ready to go? Guessing that scenario with so many tanks would mean a few extra calculations per frame.

Share this post


Link to post
Share on other sites

Testing resumes.

Nvidia Control Panel Settings

Anisotropic filtering: 16

Antialiasing - FXAA: Off

Antialiasing - Gamma Correction: On

Antialiasing - Mode: Override any application setting

Antialiasing - Setting: 16xCSAA

Antialiasing - Transparency: Multisample

Maximum Pre-rendered Frames: Application controlled

Power Management: Maximum performance

Texture Filtering - Anisotropic sample optimization: Off

Texture Filtering - Negative LOD Bias: Clamp

Texture Filtering - Quality: High quality

Texture Filtering - Trilinear Optimization: On

Threaded Optimization: Auto

Triple Buffering: Off

Texture Filtering Anisotropic Filter Optimization: Off

2013-03-13 15:01:19 - CM Normandy

Frames: 2295 - Time: 60000ms - Avg: 38.250 - Min: 26 - Max: 48

2013-03-13 15:09:19 - CM Normandy

Frames: 2319 - Time: 60000ms - Avg: 38.650 - Min: 27 - Max: 47

2013-03-13 15:10:30 - CM Normandy

Frames: 2316 - Time: 60000ms - Avg: 38.600 - Min: 27 - Max: 47

--------------------------------------------------------------------------

Nvidia Control Panel Settings

Anisotropic filtering: 16

Antialiasing - FXAA: Off

Antialiasing - Gamma Correction: On

Antialiasing - Mode: Override any application setting

Antialiasing - Setting: 16xCSAA

Antialiasing - Transparency: 8x (Supersample)

Maximum Pre-rendered Frames: Application controlled

Power Management: Maximum performance

Texture Filtering - Anisotropic sample optimization: Off

Texture Filtering - Negative LOD Bias: Clamp

Texture Filtering - Quality: High quality

Texture Filtering - Trilinear Optimization: On

Threaded Optimization: Auto

Triple Buffering: Off

Texture Filtering Anisotropic Filter Optimization: Off

2013-03-13 15:17:44 - CM Normandy

Frames: 2546 - Time: 60000ms - Avg: 42.433 - Min: 31 - Max: 56

2013-03-13 15:19:19 - CM Normandy

Frames: 2402 - Time: 60000ms - Avg: 40.033 - Min: 27 - Max: 49

2013-03-13 15:20:25 - CM Normandy

Frames: 2405 - Time: 60000ms - Avg: 40.083 - Min: 28 - Max: 49

Note: Surprise, surprise - I've discovered another inconsistency! I get 2 more average frames when setting AA transparency to 8x (Supersample) which is otherwise the most demanding level for this setting. Multisample was supposed to give more FPS's with a bit worse visual quality over any Supersample option (from 2x to 8x) but it's not so in my case. I get crystal clear visuals and no ground textures shimmering with this one - yeeeeey! Best visual experience so far with highest FPS's ever for me - this is just wow!

--------------------------------------------------------------------------

Nvidia Control Panel Settings

Anisotropic filtering: 16

Antialiasing - FXAA: Off

Antialiasing - Gamma Correction: On

Antialiasing - Mode: Override any application setting

Antialiasing - Setting: 16xCSAA

Antialiasing - Transparency: 8x (Supersample)

Maximum Pre-rendered Frames: 1

Power Management: Maximum performance

Texture Filtering - Anisotropic sample optimization: Off

Texture Filtering - Negative LOD Bias: Clamp

Texture Filtering - Quality: High quality

Texture Filtering - Trilinear Optimization: On

Threaded Optimization: Auto

Triple Buffering: Off

Texture Filtering Anisotropic Filter Optimization: Off

2013-03-13 15:59:28 - CM Normandy

Frames: 2319 - Time: 60000ms - Avg: 38.650 - Min: 26 - Max: 48

2013-03-13 16:00:36 - CM Normandy

Frames: 2369 - Time: 60000ms - Avg: 39.483 - Min: 28 - Max: 49

2013-03-13 16:05:41 - CM Normandy

Frames: 1969 - Time: 60000ms - Avg: 32.817 - Min: 20 - Max: 45

Note: I get 1-2 less frames in the first 2 tests while the 3rd one gives me a considerable drop (maybe a result of some OS process in the background) with this combo so I won't use it any more.

-------------------------------------------------------------------------------

Nvidia Control Panel Settings

Anisotropic filtering: 16

Antialiasing - FXAA: Off

Antialiasing - Gamma Correction: On

Antialiasing - Mode: Override any application setting

Antialiasing - Setting: 16xCSAA

Antialiasing - Transparency: 8x (Supersample)

Maximum Pre-rendered Frames: 4

Power Management: Maximum performance

Texture Filtering - Anisotropic sample optimization: Off

Texture Filtering - Negative LOD Bias: Clamp

Texture Filtering - Quality: High quality

Texture Filtering - Trilinear Optimization: On

Threaded Optimization: Auto

Triple Buffering: Off

Texture Filtering Anisotropic Filter Optimization: Off

2013-03-13 16:12:09 - CM Normandy

Frames: 2233 - Time: 60000ms - Avg: 37.217 - Min: 25 - Max: 47

2013-03-13 16:13:16 - CM Normandy

Frames: 2243 - Time: 60000ms - Avg: 37.383 - Min: 25 - Max: 47

2013-03-13 16:14:23 - CM Normandy

Frames: 2247 - Time: 60000ms - Avg: 37.450 - Min: 26 - Max: 47

Note: Since I'm getting roughly 3 frames drop it looks like the smartest bet is to use 3D application setting.

-------------------------------------------------------------------------------

Testing continues with the remaining nvidia control panel settings. 3 anomalies discovered so far: Nvidia control panel set to AA 16x, in-game textures set to Best and AA transparency set to 8x (Supersample) :P

Edit: I was actually using Texture Filtering - Negative LOD Bias: "Clamp" and not "Allow" - I fixed this now in the text.

Share this post


Link to post
Share on other sites

FRAPS RESULTS

2013-03-14 17:20:15 - CM Normandy

Frames: 1790 - Time: 60000ms - Avg: 29.833 - Min: 24 - Max: 34

Frames: 1790 - Time: 60000ms - Avg: 29.833 - Min: 24 - Max: 34

Frames: 1790 - Time: 60000ms - Avg: 29.833 - Min: 23 - Max: 34

SETTINGS

Screen resolution of the game:1280x768 75 hertz

In game model quality: Balanced

In game texture quality: Best

Shadows on/off: On

Shaders on/off: On

Trees (full/distant): Full

GRAPHICS CARD SETTINGS

NVIDIA Control Panel Settings

Anisotropic filtering-- 16x

Anti-aliasing-FXAA-- off

Anti-aliasing-Gamma correction-- Use global setting(on)

Anti-aliasing-Mode-- Override any application setting

Anti-aliasing-Setting-- 32x CSAA

Anti-aliasing-Transparency-- 8x (supersample)

CUDA-GPUs-- Use global setting (all)

Maximum pre-rendered frames-- 4

Multi-display/mixed-GPU acceleration-- Multiple display performance mode

Power Management mode-- Use global setting (Adaptive)

Texture filtering-Anisotropic sample optimization-- off

Texture filtering-Negative LOD bias-- Allow

Texture filtering-Quality-- High quality

Texture filtering-Trilinear optimization-- off

Threaded optimization-- off

Triple buffering--on

Texture filtering-Anisotropic filter optimization-- off

Vertical sync-- on

HARDWARE

Manufacturer + model: DIY

OS:Win 7 64 bit

CPU:Intel® Core2 Duo CPU E7500 @ 2.93 GHz

Mainboard:

RAM:4 gig

Graphics card (overclocked?):NVIDIA GeForce GTS 450 with 214.14 beta driver : no

Graphics memory:1 gig

VOLUNTARY ADDITIONAL DATA

Z folder removed.

I run a lower screen resolution so I can have the faster refresh rates to help with eye fatigue. The 60 Hz refresh rates kills my eyes and caused a headache after an hour or two.

Shadows on/off: Off

2013-03-14 17:30:12 - CM Normandy

Frames: 2474 - Time: 60000ms - Avg: 41.233 - Min: 36 - Max: 53

Frames: 2438 - Time: 60000ms - Avg: 40.633 - Min: 37 - Max: 50

Frames: 2438 - Time: 60000ms - Avg: 40.633 - Min: 37 - Max: 50

I notice that this test scenario has "overcast" skies. The shadows are really light. I wonder if brighter skies and darker shadows change the frame rates??

Resolution: 1920 1080 60

Shadows on

2013-03-14 17:43:11 - CM Normandy

Frames: 1429 - Time: 60000ms - Avg: 23.817 - Min: 19 - Max: 30

Frames: 1425 - Time: 60000ms - Avg: 23.750 - Min: 19 - Max: 29

Frames: 1424 - Time: 60000ms - Avg: 23.733 - Min: 19 - Max: 29

Share this post


Link to post
Share on other sites
@LeeW: Why don't you set it to Single display?

I did try the single display sitting, did not make any difference as the other display is off while doing CM. Need to do a test with it on I guess.:)

Share this post


Link to post
Share on other sites

Multi-display/mixed-GPU acceleration-- Multiple display performance mode

Only one monitor turned on.

2013-03-15 14:33:19 - CM Normandy

Frames: 1933 - Time: 60000ms - Avg: 32.217 - Min: 25 - Max: 39

Frames: 1941 - Time: 60000ms - Avg: 32.350 - Min: 26 - Max: 36

Both monitors turned on. CM on one and desktop on other.

2013-03-15 14:36:23 - CM Normandy

Frames: 1940 - Time: 60000ms - Avg: 32.333 - Min: 26 - Max: 36

Frames: 1937 - Time: 60000ms - Avg: 32.283 - Min: 26 - Max: 36

CM on one monitor and opened a browser window on the other.

2013-03-15 14:50:25 - CM Normandy

Frames: 2175 - Time: 60000ms - Avg: 36.250 - Min: 28 - Max: 44

Frames: 2184 - Time: 60000ms - Avg: 36.400 - Min: 30 - Max: 42

Monitors setup to duplicate at 1920 1080 60

2013-03-15 15:17:26 - CM Normandy

Frames: 1210 - Time: 60000ms - Avg: 20.167 - Min: 17 - Max: 24

Frames: 1210 - Time: 60000ms - Avg: 20.167 - Min: 17 - Max: 24

Monitors setup to duplicate at 1080 768 75

2013-03-15 15:33:51 - CM Normandy

Frames: 2189 - Time: 60000ms - Avg: 36.483 - Min: 28 - Max: 43

Frames: 2206 - Time: 60000ms - Avg: 36.767 - Min: 30 - Max: 42

Share this post


Link to post
Share on other sites
... setting AA transparency to 8x (Supersample)

I get crystal clear visuals and no ground textures shimmering with this one - yeeeeey! Best visual experience so far with highest FPS's ever for me - this is just wow!

I cannot confirm this. AA transparency 8x, 4x, 2x Supersample does not affect the shimmering of the grass.

Which seems logical to me, because AA transparency should work for transparent textures, shouldn't it?

Share this post


Link to post
Share on other sites

Results with model set to best, balanced and fastest, with and without shadows:

HW like above (CPU @2,66GHz o/ced)

SETTINGS

Screen resolution of the game:1920x1200

In game model quality: BEST

In game texture quality: BEST

Shadows on/off: ON

Shaders on/off: ON

Trees (full/distant): FULL

In game model quality: BEST

FRAPS RESULTS

13/25/19,2

In game model quality: BALANCED

FRAPS RESULTS

21/40/30,5

In game model quality: FASTEST

FRAPS RESULTS

36/58/45,1

Shadows OFF

In game model quality: BEST

FRAPS RESULTS

15/34/24,9

In game model quality: BALANCED

FRAPS RESULTS

28/53/41,8

In game model quality: FASTEST

FRAPS RESULTS

37/61/48,5

What i find interesting is the relatively low impact of shadows with the fastest model setting here (only -7% avg. framerate) compared to -27% avg.FR@model=balanced and -23% avg.FR@model=best.

Share this post


Link to post
Share on other sites
I cannot confirm this. AA transparency 8x, 4x, 2x Supersample does not affect the shimmering of the grass.

Which seems logical to me, because AA transparency should work for transparent textures, shouldn't it?

Well, I'm not making things up - in my case it changes the visual clarity and almost completely negates texture shimmering when camera is moving. Have you tried to set 8x Supersample and run the test seeing if you get any FPS gain with it in comparison to have it set to Multisample?

Share this post


Link to post
Share on other sites

Well, looks like it has to be just certain GPU's or even a combo of some hardware for one to experience what I'm experiencing.

It's so smooth and beautiful now - have no desire to upgrade my GPU any more, he he.

Share this post


Link to post
Share on other sites

You mean the Antialiasing Transparency setting set to Off and then to Supersample 8x?

Supersample 8x:

supersample8x.png

Antialiasing Transparency Off:

antialiasingtransparenc.png

It's hard to see shimmering differences on any still pic since shimmering by default happens when camera moves around. Still you can spot that Supersample 8x has a nicer look to it. Take note especially to trees in the distance. The ones with the setting on Off are more grainy. Same applies to ground grass texture.

Edit: Hmm, it's harder to see it here on the forum since SS's are not in the original size...

Share this post


Link to post
Share on other sites

I believe i can see it. :D Seems to look a bit less grainy.

But a big difference?

ps: since you seem to like mods, there is a great tree mod with less leaves. Looks better, more realistic and increases the visibility of units behind trees.

Share this post


Link to post
Share on other sites

Yes, it makes a big difference because I get no shimmering and this special crystal clear visuals still pic can't show you unfortunately.

Ah I remember that mod - will apply it (forgot to do so after I upgraded to 2.0).

Edit: It's EZ's Less Leaves mod right?

Share this post


Link to post
Share on other sites

Hister,

After looking at your picture, I decided my test weren't run correctly. I had always left Barkmann's unit selected when running the test. (green icon under the unit) I ran a test to confirm that this would make a difference. This test is in Iron mode, so it makes a big difference in what icons and units are visible when running the test.

All settings as per the original test for me.

Balanced Barkmann selected

2013-03-19 20:19:45 - CM Normandy

Frames: 2140 - Time: 60000ms - Avg: 35.667 - Min: 28 - Max: 43

Frames: 2140 - Time: 60000ms - Avg: 35.667 - Min: 29 - Max: 41

Frames: 2145 - Time: 60000ms - Avg: 35.750 - Min: 29 - Max: 41

Balanced Barkmann unselected (right click map)

2013-03-19 20:23:38 - CM Normandy

Frames: 1795 - Time: 60000ms - Avg: 29.917 - Min: 25 - Max: 34

Frames: 1806 - Time: 60000ms - Avg: 30.100 - Min: 25 - Max: 34

Frames: 1812 - Time: 60000ms - Avg: 30.200 - Min: 25 - Max: 34

Share this post


Link to post
Share on other sites

Two counteríntuitive performance issues (to me anyway):

Editing a large or huge map in the editor puts quite a strain on the PC as evidenced by chugging scrolling. Shouldn´t the map size be irrelevant in the editor? it´s just little colored squares, not 3D calculations.

The other day I set up a large QB battle on a medium map with lots of infantry. While setting up the troops, the FPS go down to practically nothing (<5) making the process quite painful. Once the game starts however, FPS are fine (20somefink).

Share this post


Link to post
Share on other sites

×