10% fps increase only by editing file terrainoptions41.lua

I noticed in latest 2.5.6.50979 open beta a parameter which hurts fps near 10% or so: concretely, in file terrainoptions41.lua.

In that file, there is a line which saids:

multiAppendBufferSize = 524288;

If you put:

multiAppendBufferSize = 400000; you will notice a 10% increase in FPS (at least, it seems to me an increase). In fact, I look old beta&stable versions and the parameter by default were 400000. It seems that new parameter of 524288 cause some fps issues.

Devs seriously, look more in detail about optimization, this is a mess for users every time editing files.

Hi Gryzor,

did you test it with all maps? Would be interesting if they all still work then.

Anyway, good find, and I will try out.

I just would not be that unfriendly to the devs, this setting might have reasons you currently don't know, dont you?

But in the current situation I understand that most people are getting mad about ED, one sometimes comes to the conclusion it seems there is nobody testing anything :noexpression:

Its getting better, slowly.

I noticed in latest 2.5.6.50979 open beta a parameter which hurts fps near 10% or so: concretely, in file terrainoptions41.lua.

 

In that file, there is a line which saids:

multiAppendBufferSize = 524288;

 

If you put:

multiAppendBufferSize = 400000; you will notice a 10% increase in FPS (at least, it seems to me an increase). In fact, I look old beta&stable versions and the parameter by default were 400000. It seems that new parameter of 524288 cause some fps issues.

 

Devs seriously, look more in detail about optimization, this is a mess for users every time editing files.

Wow! How on earth did you manage to find this? :thumbup:

I'll try it out tomorrow when I get the chance

In the meanwhile, can anyone confirm that this is working?

Hi Gryzor,

 

 

did you test it with all maps? Would be interesting if they all still work then.

Anyway, good find, and I will try out.

 

 

I just would not be that unfriendly to the devs, this setting might have reasons you currently don't know, dont you?

 

But in the current situation I understand that most people are getting mad about ED, one sometimes comes to the conclusion it seems there is nobody testing anything :noexpression:

Its getting better, slowly.

I tested in Dubai when taking off with many building around: With 400.000 I have fps in VR 45 stable (motion smoothing enabled). With 524288, It fluctuates between 45 and 30 fps.

In other terrain, caucasus, it appears less stutters when TV cameras or IR sensors are enabled (by pointing them at the horizont, seems a bit smoothier).

Wow! How on earth did you manage to find this? :thumbup:

I'll try it out tomorrow when I get the chance

 

In the meanwhile, can anyone confirm that this is working?

By comparing with some older config files which I made some backups, and I suspected that some sort of config files could hurt fps or performance (like terrainoptions41.lue). I noticed the older parameter of 400.000 was changed from the latest versions, just when ground radar were introduced due F-18 improvements.

Sticky so I can find this!

I've had the multiAppendBufferSize = 400000; included in terrainoptions41.lua of the PG mod pack for a long time.

https://forums.eagle.ru/showthread.php?t=211165

Mod terrainoptions41.lua is not intended for the Caucasus?

Mod terrainoptions41.lua is not intended for the Caucasus?

terrainoptions41.lua settings are apply to ALL terrains

Thank you for the clarification.

.

.

With 400.000 I have fps in VR 45 stable (motion smoothing enabled).

.

With 524288, It fluctuates between 45 and 30 fps.

.

.

Would seem like a 33% boost, not 10% :smartass:

Subbed to this thread for future reference/testing.

I tried it but did not notice any improvement. I think I am memory-speed bound (down-clocking it reduces FPS linearly). Shaders Mod, on contrary, has no effect. I decided to wait for next-gen DDR5 CPU to upgrade all at once. I am usually 45 FPS in the air with F18, but 20 on SC and in the missions with 3 wingmen/many AI. My preload radius is 70000, visibility Medium. Anything lower and I can't see ground targets.

Could somebody argue if upgrading my DDR3 memory size to 32Gb will work as an interim solution to improve FPS in busy missions? I'd need to replace it all, would hate if it does not make a difference...

Be aware that lowering this number will cause momentary objects to disappear/reappear out of view, this might be why the FPS increase is there since objects are being culled - if you want to really see this in action drop that number lower to around 100000 or less.

Be aware that lowering this number will cause momentary objects to disappear/reappear out of view, this might be why the FPS increase is there since objects are being culled - if you want to really see this in action drop that number lower to around 100000 or less.

Thanks for sharing.

It is probably rendering something less to get more FPS, IMO. I don't think it's for free.

It is probably rendering something less to get more FPS, IMO. I don't think it's for free.

Correct, as I mentioned above the lower the setting the more aggressive the object culling is resulting in greater FPS - ED probably increased this recently as there may have been times where the pop-in would have occured so increasing this solves it.

524288/1024 = 512

1 Mbyte = 1024 kbits

Its Probably about memory size of some texture layer tile...

Decreasing the value decreases visibility, but give the framerate performance, have anyone tried to compare other values. What role in it plays the preload radius?

524288/1024 = 512

 

1 Mbyte = 1024 kbits

 

Its Probably about memory size of some texture layer tile...

No, it's a buffersize for Geometry Instancer - the number of objects rendered within your FOV, the lower the number the more objects get culled, the greater the number the more objects are allowed to be drawn - this is actually very useful for pushing up the distancFactor parameter in the graphics.lua, with a very high number I can use a setting of 3.0 or more with zero objects in scene disappearing and reappearing, pretty awesome :)

I dont notice any difference with objects disappearing or quality between 524288 and 400000.

In fact, with 524288 in VR you notice how some objects are not rendered in one eye (6 month bug since 2.5.6).

However fps are a bit higher with 400000, which is specially important in VR.

I tried it but did not notice any improvement. I think I am memory-speed bound (down-clocking it reduces FPS linearly).

Im in the same boat with DDR4-3200. FPS scales 100% linearly with ram speed, something very few synthetic benchmarks even do. All these people buying the fastest CPUs out there and overclocking them thinking every MHz counts, I wonder if the more important reason they are getting performance boost is that they are also overclocking their memory controllers and cache.

Hi,

I made a few synthetic benchmark test -time 02:54 min. Only then the increase of the parameter proposed by the ED makes more sense.

Changing the value will achieve a higher framerate, unfortunately at the expense of those of the minimum, which are for fluency more important.

After adjustment removed FXO/metashaders2.

- Setting preset High

- SSLR = Off

- Mirror = Off

- Resolution = 2560 x 1440

MultiAppendBufferSize = 524288

22-06-2020, 11:37:29 DCS.exe benchmark completed, 13421 frames rendered in 172.350 s

Average framerate : 77.8 FPS

Minimum framerate : 51.2 FPS

Maximum framerate : 105.1 FPS

1% low framerate : 51.3 FPS

0.1% low framerate : 37.0 FPS

MultiAppendBufferSize = 400000

22-06-2020, 11:47:13 DCS.exe benchmark completed, 13898 frames rendered in 173.067 s

Average framerate : 80.3 FPS +3,21 %

Minimum framerate : 53.9 FPS +5,27 %

Maximum framerate : 114.5 FPS +8,94 %

1% low framerate : 40.5 FPS -26,67 %

0.1% low framerate : 30.0 FPS -23,33 %

Benchmark256new2.trk

Hi,

I made a few synthetic benchmark test -time 02:54 min. Only then the increase of the parameter proposed by the ED makes more sense.

Changing the value will achieve a higher framerate, unfortunately at the expense of those of the minimum, which are for fluency more important.

 

After adjustment removed FXO/metashaders2.

 

- Setting preset High

- SSLR = Off

- Mirror = Off

- Resolution = 2560 x 1440

 

MultiAppendBufferSize = 524288

22-06-2020, 11:37:29 DCS.exe benchmark completed, 13421 frames rendered in 172.350 s

Average framerate : 77.8 FPS

Minimum framerate : 51.2 FPS

Maximum framerate : 105.1 FPS

1% low framerate : 51.3 FPS

0.1% low framerate : 37.0 FPS

 

MultiAppendBufferSize = 400000

22-06-2020, 11:47:13 DCS.exe benchmark completed, 13898 frames rendered in 173.067 s

Average framerate : 80.3 FPS +3,21 %

Minimum framerate : 53.9 FPS +5,27 %

Maximum framerate : 114.5 FPS +8,94 %

1% low framerate : 40.5 FPS -26,67 %

0.1% low framerate : 30.0 FPS -23,33 %

It's a change... :D

I'll test it tonight on my VR rig. :D

Probarbly boost some FPS. We'll see....

CONTINUE:

Tested. No significant changes. I set it back, thanks for the idea.

Interesting. Kinda makes sense whatever is being buffered could reduce stuttering if you buffer more of it. But how repeatable is that benchmark? if you run it a second time, what kind of variation are you seeing? Ive found playing back tracks can give me very significant variations between runs, even changing nothing.