Horrible frames. Getting frustrated.

[Howler]

Back to CMSF2/ReShade 4.1, stationary camera (#2).

Fastest 28 FPS (~ 35ms)

Balanced 28 FPS (~ 35ms)

Improved 25 FPS (~ 40ms)

Best  17 FPS (~ 60ms)

Running again with ReShade off yielded:

Fastest 31 FPS (~30ms)

No Change to Balanced or Improved

Best 20 FPS (~ 50ms)

Above was run by changing 3D Model setting in game and rewinding the turn. I believe slightly better results would be had if one was to set the quality in the menu and then loading a save...

I'll try to determine the size of detail bubble next. Again, I'm consistently seeing significant drop between 'balanced' and 'best'.

[Howler]

Last post today on this matter. Promise.

Bubble Best/Balanced/Fast:

Camera #2 locked on unit - 600m/240m/100m

Camera #3 locked on unit - 340m/200m/220m

Camera #5 locked on unit - 50m/0m/0m

Turning around a locked unit using the rotate hotkey (eg 'Q') a couple of times as Best/Balanced/Fast in FPS:

Camera #2 - 25/45/40

Camera #3 - 20/30/35

Camera #5 - 30/45/60

In all cases, holding down the Rotate Left key ('Q') for more than a couple of complete turns resulted in the FPS counter shooting up which seems to indicate caching...

The bubble presented as a cone and I opted to measure the shortest distance. Save game used was 'Factory Outlet'. Finally, the ReShade FPS counter was used as AMD does not provide one.

More importantly, the response is smooth. Zoom and camera movement is immediate. Turning off the AA in-game and in the video driver has eliminated any lag.

[37mm]

5 hours ago, Howler said:

 Further, changing the model detail while in game skews the number. I've had faster renders at every setting when you set the detail in the option menu then load the save game.

I very much doubt you'll be regularly getting 300-400m bubbles playing on giant complex maps like the Bridgehead... not unless one has an overclocked, liquid nitrogen cooled CPU (CM IS CPU intensive not really graphics card intensive, pure Gigahertz aids the game more than anything).

For such maps, it's a matter of finding that happy balance between a sufficient "visual bubble" for your play & your fps... playing on 3d Model Quality 'best' was always likely going to be a struggle on such a map.

I'm glad that I'm not the only one who's found turning off all AA & reducing your 3d Model quality can be of assistance to smooth play.

 

BTW

You can use Reshade to "reload" a scene from scratch so you don't have to go back to the option menu everytime... I suppose you could also alt-tab & click back to CM.

[Howler]

1 hour ago, 37mm said:

For such maps, it's a matter of finding that happy balance between a sufficient "visual bubble" for your play & your fps... playing on 3d Model Quality 'best' was always likely going to be a struggle on such a map.

Agreed. You've got me now trying 'balanced' to see if the larger maps become playable.

1 hour ago, 37mm said:

You can use Reshade to "reload" a scene from scratch so you don't have to go back to the option menu everytime... I suppose you could also alt-tab & click back to CM.

Any comments I made regarding the option menu were based on my assumption that setting the 3D model in the options, as opposed to the hotkey in-game, may produce more accurate FPS measurement. One then starts or loads a game without concern for stale buffers/caches.

In any event, my numbers were measured first by using the hotkey and then rewinding the turn. I therefore loaded a save game once and never exited the game between measurements for the various 3D model settings.

[AstroCat]

We can only hope one day, we get a graphics engine even at this level. I've upgraded to a i7-9700K@4.9GHz and an EVGA RTX 2080 Ti XC Ultra and still get terrible fps! Talking hitting 20fps and below on just a simple medium hills quick battle, not even with many trees. Super sluggish and laggy camera movement. lol!

IL-2 Sturmovik - Tank Crew

https://www.youtube.com/watch?v=fwkTxyh21-g
https://www.youtube.com/watch?v=3tQVifFTGKY
https://www.facebook.com/il2sturmovik/posts/2026223027455545?__tn__=K-R

 

 

Edited February 22, 2019 by AstroCat

[RockinHarry]

26 minutes ago, AstroCat said:

We can only hope one day, we get a graphics engine even at this level. I've upgraded to a i7-9700K@4.9GHz and an EVGA RTX 2080 Ti XC Ultra and still get terrible fps! Talking hitting 20fps and below on just a simple medium hills quick battle, not even with many trees. Super sluggish and laggy camera movement. lol!

IL-2 Sturmovik - Tank Crew

https://www.youtube.com/watch?v=fwkTxyh21-g
https://www.youtube.com/watch?v=3tQVifFTGKY
https://www.facebook.com/il2sturmovik/posts/2026223027455545?__tn__=K-R

 

 

thanks for posting! Didn´t yet have that on the radar! Looks very very promising. 

http://simhq.com/forum/ubbthreads.php/topics/4435032/tank-crew-clash-at-prokhorovka

[37mm]

51 minutes ago, AstroCat said:

We can only hope one day, we get a graphics engine even at this level. I've upgraded to a i7-9700K@4.9GHz and an EVGA RTX 2080 Ti XC Ultra and still get terrible fps! Talking hitting 20fps and below on just a simple medium hills quick battle, not even with many trees. Super sluggish and laggy camera movement. lol!

 

You have twice the CPU power I have & when I first started with CMSF2 I was running at ~10-15fps.

I now regularly play at 40-60 fps on the type of maps you describe (and that's very often at an 1800p resolution) so almost certainly, with a bit of work, we can get your fps up & improve your playing experience.

First things first... have you read & tried out DeutschRitters excellent advice?

Secondly, have you tried turning off ALL AA (turning off Vsync also seems to help)... both on your graphics cards settings & ingame options?

 

Edited February 22, 2019 by 37mm

[AstroCat]

Thanks for the reply, I've done so much testing and tweaking through the years it's nuts. One thing though, and it can at times make a big fps difference but I can not play these games without SS/AA, otherwise it becomes a shimmering texture crawl mess that legit hurts my eyes to look at.

Edited February 22, 2019 by AstroCat

[AstroCat]

1 hour ago, RockinHarry said:

thanks for posting! Didn´t yet have that on the radar! Looks very very promising. 

http://simhq.com/forum/ubbthreads.php/topics/4435032/tank-crew-clash-at-prokhorovka

Yeah it's the full ground extension of the really great flight sim IL2. 

[RockinHarry]

6 minutes ago, AstroCat said:

Yeah it's the full ground extension of the really great flight sim IL2. 

yep, Il2 was always my favourite flight sim and they also succeeded in making the ground looking great from the skies. I´ve great expectations of tank crew and if they´ll yield some good money they might extend to combined arms further on.  But multiplayer crew in realistically modelled WW2 tanks might take me to online multiplayer games at last.

[37mm]

22 minutes ago, AstroCat said:

Thanks for the reply, I've done so much testing and tweaking through the years it's nuts. One thing though, and it can at times make a big fps difference but I can not play these games without SS/AA, otherwise it becomes a shimmering texture crawl mess that legit hurts my eyes to look at.

I understand the need for SS/AA... we'll get to that eventually.

At the moment I'm mostly concerned about your fps, let's get that up first.

Have you set your 3d Model Quality to balanced & Vsync on FAST (or even off) as per DeutschRitters advice? How is the ingame fps?

Likewise, how is your FPS with ALL AA turn off?

I reiterate that I had similiar FPS (worse even) issues to yourself just four months ago, yet this was me last night on a system much, much worse than your own...

 

 

[OlafP]

OpenGL might be bad, but it's not going to hold back your fps like that in a game that has very simplistic graphics. And game calculations are not the culprit, because when you move the camera off map, and frustum culling kicks in, the fps skyrockets. When you move the camera on map again, the fps is back to its low rate. Hence, the poor performance is caused by the rendered geometry, not game calculations. It wouldn't make any sense to do game calculations in the orders phase anyway, because those calculations would be dependent on orders that aren't yet given.

When you position the camera with only a slight portion of the map exposed, something interesting happens:

This is on a GTX 1080, I have nothing running in the background and there is no input from me. The game engine is choking that graphics card while producing nothing but a few trees and a small patch of ground!

Notice three things:

- The framerate fluctuates from very low (around 7 fps) to very high (approaching 300 fps). 

- The framerate eventually stabilizes at a constant fps that is relatively low (43 at the beginning of the video, 54 at the end).

- When the trees and ground are rendered in high detail, the framerate is high, while it is low or lower when they are rendered in low detail. The only exception to this is when the framerate is stable at either 43 or 54.

So from the last point, we can deduce that for that given perspective, I should be able to have very high fps AND high detail at the same time. 

What I think is going on here is that the game engine is constantly trying to reduce or increase the level of detail based on current performance and current level of detail. It does this by going into a negative feedback loop which eventually stabilizes at an equilibrium based on the given graphics setting (fast; faster; balanced etc.). The only problem is that this mechanism itself seems to be taking up a lot more GPU power than whatever small amount is required for the rendering. The consequent reduction in performance, caused by the LOD simplifier, is then fed back into the feedback loop as input, causing the whole system to run the game at an fps rating and detail that has nothing to do with available system performance.

I used to play this game, or engine, on an old laptop back in 2012. The game performed comparable to what it does on my GTX 1080 today. Clearly, this doesn't make any sense. What seems to be going on is a self-defeating system that holds back performance on modern computers (but might possibly function properly on laptops or low-end systems). 

[37mm]

1 hour ago, OlafP said:

What I think is going on here is that the game engine is constantly trying to reduce or increase the level of detail based on current performance and current level of detail. It does this by going into a negative feedback loop which eventually stabilizes at an equilibrium based on the given graphics setting (fast; faster; balanced etc.). The only problem is that this mechanism itself seems to be taking up a lot more GPU power than whatever small amount is required for the rendering. The consequent reduction in performance, caused by the LOD simplifier, is then fed back into the feedback loop as input, causing the whole system to run the game at an fps rating and detail that has nothing to do with available system performance.

I used to play this game, or engine, on an old laptop back in 2012. The game performed comparable to what it does on my GTX 1080 today. Clearly, this doesn't make any sense. What seems to be going on is a self-defeating system that holds back performance on modern computers (but might possibly function properly on laptops or low-end systems). 

I would've agreed with your conclusions except for one issue... I cannot recreate the effect.

I can recreate the "frame rate fluctuations that stabilize into an equilibrium" effect easily enough because, as you suggest, that's happening pretty much all the time.

However the fluctuations I have (no matter the settings I used... AA on or off, Best/Balanced 3d model quality, high or low tree details) are only a few fps here or there... even on the edge of a forested map.

Obviously, the fluctuations are greater when I pan & move the camera... however 10-20 fps is the usual level of fluctuation not hundreds of fps like you're getting.

I can't seem to recreate these wild fluctuations.

I'd suggest that you have some kind of driver issue?

What happens when you have vsync on Fast & recreate that scene?

[OlafP]

1 hour ago, 37mm said:

I would've agreed with your conclusions except for one issue... I cannot recreate the effect.

I can recreate the "frame rate fluctuations that stabilize into an equilibrium" effect easily enough because, as you suggest, that's happening pretty much all the time.

However the fluctuations I have (no matter the settings I used... AA on or off, Best/Balanced 3d model quality, high or low tree details) are only a few fps here or there... even on the edge of a forested map.

Obviously, the fluctuations are greater when I pan & move the camera... however 10-20 fps is the usual level of fluctuation not hundreds of fps like you're getting.

I can't seem to recreate these wild fluctuations.

I'd suggest that you have some kind of driver issue?

What happens when you have vsync on Fast & recreate that scene?

The issue is not the fluctuation itself, although that is pretty annoying too. The issue is that you end up with terrible fps whenever the fluctuations stabilize. The fluctuations would be a lot less wild on a less powerful graphics card, since the system would not be able to take the fps so high. That could possibly explain why you are not seeing them.

V-sync doesn't make any difference.

[37mm]

11 minutes ago, OlafP said:

The issue is not the fluctuation itself, although that is pretty annoying too. The issue is that you end up with terrible fps whenever the fluctuations stabilize. The fluctuations would be a lot less wild on a less powerful graphics card, since the system would not be able to take the fps so high. That could possibly explain why you are not seeing them.

V-sync doesn't make any difference.

Ignoring whether or not 50 fps should be considered terrible for CM perhaps we could test this theory more thoroughly?

Play the game with your PC on minimal power settings... that should turn your high end computer into a low end one.

Any difference?

 

When you say V-sync make's no difference... do you mean it still stabilizes at 40-55 fps or that it still fluctuates up to 300fps? If the latter then vsync can't be working.