Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[DougW60]

I have a Radeon 7970, 27in Monitor (2048x1152 resolution), 12xgig 1600 RAM and a I7 960 CPU.  The CPU, RAM and monitor are all 6 years old.  I have the same problem as identified by edc1 with FPS at approximately 60 plus at altitude but when I accelerate time speed to 4x and 8x my framerates dropped dramatically.  The point is, edc1 is using an nVidia GPU and I'm a AMD GPU, so I feel confident that the problems described are not tied to the GPU. We have a bottleneck in our system, no big deal, I am upgrading to a new CPU/motherboard/RAM within the next couple of months (I7 6700K) and once AMD and nVidia bring a single GPU that can handle 4k resolution, I'll upgrade to 4k.

[sniperton]

I have the same problem ... when I accelerate time speed to 4x and 8x my framerates dropped dramatically. ... I feel confident that the problems described are not tied to the GPU.

You can make it sure by reducing screen resolution. If you still have a dramatic FPS drop then it's clearly not related to your GPU.

We have a bottleneck in our system

Sure we have, but the problem seems to be originating in some unoptimized (or conflicting) program code overburdening the CPU. In contrast, in HSFX there's no significant FPS drop when I accelerate time. (BTW, I've read somewhere that enabling the WAW_FLAME mod via JSGME might eliminate the problem.)

[Chupacabras84]

Curious thing I noticed.
It needs further testing but for quite a while I felt that certain aircraft limit available air.ini size more than the others.
For quite a while I suspected that duplicates are the ones that doing it for example J7W1a and J7W1b.
I am currently making pacific theatre and I hit fps block, I am deleting 4 random planes and I am not gaining a single fps but its enough to disable either of those planes and fps goes up 10 frames.
Earlier I've got the feeling that different version of Me 262 did that or different variation of corsair.
You can disable 10 other planes but fps wont change but its enough to delete just one or two dupes and framerate goes up.

Maybe different planes accessing same FM somehow causing it or there is conflict of some other kind.
Unfortunately I cant look into it deeper than just scratching the surface 

[edc1]

Hi all,
just started playing Hawaii 41 campaign, and increased the speed x8 (to see if the FPS drop was the same)
But it only went from 60 to 58! Not 60 to 4.
So the drop must be down to either :- 1 the map                                                         
                                                                   2 the aircraft or number of aircraft (which I do not belive in my instant)
                                                                   3 I am asking the program to do too much in my mission building
The map used in Hawaii 41is not as detailed as the map (ItalyAfricaGreece which gives me this massive FPS
drop down to 4 from 60.
I now think the only people to throw some light on the problem of FPS drop when speeding up a mission  are those who know and understand programming.
I know this won't help but maybe it's another piece to the jigsaw.
Cheers edc1 (stan)

[sniperton]

I am deleting 4 random planes and I am not gaining a single fps but its enough to disable either of those planes and fps goes up 10 frames.
...
Unfortunately I cant look into it deeper than just scratching the surface.

Great find. I suggest you should make it a new thread and ask for help from modders who know how various models are crosslinked in-game.

[Chupacabras84]

I might try and ask someone later when I finish figuring this out a bit more, for example F4U-5 with N variant also have similar effect but even if I remove all of them with full WaW air.ini my fps wont go up very high, framerate needs to be already in the ballpark but yeah, certain planes weight much more.

Its hard to say whether there is conflict or FM for those planes is just too heavy, I have no idea how FM's are but I would imagine its a table of some kind, maybe there is something about those FM's that causing it, maybe modders went overboard with resolution.
Assuming its the FM if there is anything unusual about it maybe it would allow to find other planes like this in WaW and optimise them for performance but as you can see, thats a lot of 'maybe' 

[decipher]

in another thread about stuttering i found this reply from wolfiz that fits to the discussion perfectly and sheds at least some more more light on this topic:

The number of air.ini entries does affect performance significantly. as every plane's .emd is loading into memory on game start, regardless of whether it is used or not.

Apparently memory reserved for terrain generation is affected past a certain number in air.ini, causing a lot of memory swapping and stuttering, especially when zooming in toward the horizon, my fps was dropping to 1-2 fps on full zoom using FOV mod.

The other issue, performance when viewing dense city .msh, is simply a matter of raw CPU power, MB bus speed and MB memory speed. I have been working a little on a simple workaround that replaces some of the static building msh with the smallest static object msh in IL2, which now renders invisible, some of the buildings are gone, but factory areas are complete with a longer range LOD, and FPS is much better.

I plan to work on that a little more the next few weeks, and see if it is practical to eventually put together a small upload package.

link to original posting: https://www.sas1946.com/main/index.php/topic,45822.msg521710.html#msg521710

if one could eventually "hack" the actual il-2 (java?) engine core and for example fix mod it to only load specifically those EMD files (FM files?) for aircraft that are actually used in the playing mission... freeing up resources for the map and ground objects etc... *dreaming*

[Chupacabras84]

Well, after finally unpacking buttons after all those years of mystery and looking at what we have here, everything looks simpler than I expected it to.
Just parameters and a value, the only thing that looks like it has any resolution is calculations of either lift surface or drag (at least I am guessing thats what that is).
In any case, I doubt IL2 would touch those values when aircraft is not in use so I guess maybe its not heavy FM after all.
I noticed some emd files pack quite a few of engine types, maybe planes that load emd in which there are many engine types are heavier than those that point to one engine type onle.

[SAS~Storebror]

The number of air.ini entries does affect performance significantly.

Yes and no.
Up to a specific point, you can load as many aircraft in parallel as you wish.
When that point is reached, the next loaded aircraft will kill your FPS.
The problem is: It depends on the user system where or when this happens. A number "x" of aircraft might still run fine on system "A" while it already freaks out FPS on system "B".

as every plane's .emd is loading into memory on game start, regardless of whether it is used or not.

Ah... no. It's not related to planes' flight- or engine models at all, it's a pure java issue.

Well, after finally unpacking buttons after all those years of mystery

Mystery? What?

maybe planes that load emd in which there are many engine types are heavier than those that point to one engine type only.

No. As stated above, it's not related to fmd or emd files at all.

Best regards - Mike

[Chupacabras84]

Mystery? What?

For years discussion on how to unpack buttons file and play with FM was banned on pretty much every IL2 forum I ever visited.
Even this forum had a note that discussion is not allowed.

Well, from the look of it, the solution is beyond my reach.
Thanks for clarification Mike.

[sniperton]

Many thanks to everyone who took the pain to investigate, comment on, and clarify this issue. Please correct me if I'm wrong:

Number of aircraft in the package (here: WAW) only affects initial loading speed of the game.
Game performance is affected by the number of activated aircraft only. Those which are remed out in the air.ini do not count, as they are not loaded.
Each aircraft entry counts only once, contrary to what Chupa hypothesised, and it doesn't matter whether it's a duplicate or something of the kind. I understand that it's a pure Java issue which affects all game versions.
Up to an N number of loaded/activated aircraft the game performance is not affected. N depends on the individual PC, being lower on weaker rigs and higher on stronger rigs. Motherboard speed and CPU power are the main factors which limit how many planes one can have without FPS loss.
Based on HSFX, I can say that 580 entries in the air.ini are still okay for a weak PC with a single thread performance benchmark of 890. Chupa's solution with reduced 'period' and 'theatre' air.inis should theoretically make WAW playable for all.

What I still don't understand:

I have a weak GPU, so that I have to reduce screen resolution to be able to play the stock game and HSFX with maxed-out graphic settings. I have decent framerates there, never dropping below 25.
The same trick doesn't work for WAW. Tuning down the graphics further on doesn't help considerably, even with an air.ini reduced to 520 entries. Behind the air.ini issue there seems to be another one, possibly related to particle effects, maps, textures, and objects.

Is there any theoretic chance that reming out maps and objects would bring back FPS the same way as it does with aircraft?

[Mick]

... Behind the air.ini issue there seems to be another one, possibly related to particle effects, maps, textures, and objects.

... then a little tweaking in the conf.ini may help ...

... like :

[Render_OpenGL]
TexQual=3
TexMipFilter=3
TexCompress=2
TexFlags.UseDither=0
TexFlags.UseAlpha=0
TexFlags.UseIndex=1
TexFlags.PolygonStipple=0
TexFlags.UseClampedSprites=1
TexFlags.DrawLandByTriangles=0
TexFlags.UseVertexArrays=1
TexFlags.DisableAPIExtensions=0
TexFlags.ARBMultitextureExt=1
TexFlags.TexEnvCombineExt=1
TexFlags.SecondaryColorExt=1
TexFlags.VertexArrayExt=1
TexFlags.ClipHintExt=1
TexFlags.UsePaletteExt=0
TexFlags.TexAnisotropicExt=1
TexFlags.TexCompressARBExt=1