Performance Test III - pr_fpstest3

[AfterDune]

Yet another attempt at finding the cause of the fps drops.

This time, we're still loading all of PR's files, but I've added 4 vanilla BF2 maps and replaced all assets with PR stuff, so PR weapons, vehicles, factions, effects, etc. Also increased view distance, overgrowth draw distance, fog range, etc.

When I testrun this, FPS goes through the roof, it's awesome! So the exact opposite of what we encounter with PR maps...

The big question is: why?


Download
Link: https://dl.dropboxusercontent.com/u/104 ... stest3.rar (~8MB)


How do I run this?
The rar contains a folder "pr_fpstest3". Place this folder inside the mods-directory of BF2. In most cases, this is something like:

Windows XP: C:\Program Files\EA Games\Battlefield 2\mods\
Windows Vista/7/8: C:\Program Files (x86)\EA Games\Battlefield 2\mods\

You can run "pr_fpstest3" using PRLauncher.exe, located here:
32-bit: bf2-dir/mods/pr_fpstest3/bin/bin32/PRLauncher.exe
64-bit: bf2-dir/mods/pr_fpstest3/bin/bin64/PRLauncher.exe


So, what do we need to do?
Well, test it out, see what results you get.

And then, we should start looking at replacing vehicles or statics with PR statics or deployables. It doesn't matter if a map looks **** because of this, it's just trying things out. You have to have a tiny bit of modding knowledge though...


Help each other out
I have no time left today, so if someone wants to write a tiny tut on how to replace vehicles and statics, that would be great. Help each other out here and share your findings.



Thanks!

[Mineral]

This is great work AD! I'll test it out later tonight.

[Mats391]

Did first small test with Karkand. These are the average fps i got.

Code: Select all

Vanilla: 100+
2xviewdistance (1000): 100+
2xstatics (2672): 65
2xVD+2xstatics: 60

To double the statics i just copied every static so each static exists twice at same position. The 2672 are only roughly 1/4 of what Fallujah uses.
I also noticed that the flames of the ambient fire were moving very fast.

[sylent/shooter]

• when walking through dust animations from bullets hitting the ground. FPS dips by 30 (from 99/10 - 65*)

^ regarding the above post. I have made a video recording this finding. When shooting at the ground with every kit from different forces. The issue persists. However, if you look away from the area that was shot FPS increases to normal levels. But when you look back, even after the bullets have stop hitting the ground after 1 second, the issue still persists.

This seems to only be on sand materials as the animation for the effects is fairly long. I did not notice it AS MUCH when shooting rock materials.

I'll keep updating as I find more issues.

[SANGUE-RUIM]

will edit here and post results later... thx

[Daniel]

First of all, takes AGES till game menu comes after the black screen...
2nd I noticed was on Mashtuur, Anti-aliasing was quite low, THOUGH I have set it to 8x ! ! In PR everyhing is MUCH sharper.

So, something is defenitely wrong here already with anti-aliasing...

Btw I have 100 fps on Mashtuur, everything seems fine to me. Couldn't replace any objects, dunno how to.

[Daniel]

Ok, the loading JUST took longer than normal since I was watching YouTube at the same time on my 2nd screen.


HOW CAN YOU set AA to 32x in PR ??? O_O IT IS BF2, there is no option...

[saXoni]

HOW CAN YOU set AA to 32x in PR ??? O_O IT IS BF2, there is no option...

Via the NVIDIA control panel.

[ShockUnitBlack]

Or CCC, if you're using a Radeon card.

[Arab]

I think I found out the root problem, as noted on the 'UAV causes lag' thread I made. It's the UAV on certain maps that causes it to lag, and when the UAV isn't used, it's back to normal.

Which maps do people get lags? Report people

I'll put my reports:
-Black Gold Insurgency
-Burning Sands Standard AAS

[iwingi]

I think I found out the root problem, as noted on the 'UAV causes lag' thread I made. It's the UAV on certain maps that causes it to lag, and when the UAV isn't used, it's back to normal.

Which maps do people get lags? Report people

I'll put my reports:
-Black Gold Insurgency
-Burning Sands Standard AAS

But, still shooting the ground makes your FPS drop like a M*t**r F**k*r!

[Arab]

But, still shooting the ground makes your FPS drop like a M*t**r F**k*r!

I don't have that though, FPS runs without it lagging and it's not noticeable at-least for me. I guess it depends on how fairly recent your graphics card is. Something like Radeon 4000+ would do good since one dev at the 'Ask a dev a question' thread had warned to upgrade the computer as it's a bit more resource intensive.

In Muttrah City AAS where the Huey was deploying flares as it was making a sharp manuver ne from East City.

Yeah, not really sure how you could pin point the behavior, but the new FPSTest shall narrow things down to particular things, if the developers already had a list to cross off of possible reasons.

[[FSA]IrRahman]

I am 80% sure that FPS drop is caused by new effects. New explosions, flares, splashes when shooting ground, fire etc. There should be option to switch versions from 'old' (0.9 to 'new' (1.0) for every each of these effects, or add 'ultra' settings for 1.0 effects and lighting.

[AfterDune]

"There should be an option to switch version"

It's really not as simple as that . I could create an fpstest4, where I overwrite all effects with "nothing", so there are no visible effects whatsoever. Sure that will increase fps, but people have also been mentioning having fps drops when nothing was going on whatsoever - no firing, no moving objects, nothing. Just, standing there, looking at a city for example. Unsure if "switching back to the old effects" is going to helps us that much there.

It's worth a shot, but still.

Btw, does anyone still have 0.98 on his pc? Perhaps you can try to run 0.98 and shoot at the ground, see what that does to your fps. We can't just shout "it's the effects!" when we have nothing* to compare it to.

(*) I'm talking about doing the same tests in 0.98 as we do in 1.0

[Spush]

Should really look into optimizing our meshes, and textures. Reduce draw calls (which are textures sheets that eat up GPU performance the most), and better lods.

[Mats391]

Btw, does anyone still have 0.98 on his pc? Perhaps you can try to run 0.98 and shoot at the ground, see what that does to your fps. We can't just shout "it's the effects!" when we have nothing* to compare it to.

(*) I'm talking about doing the same tests in 0.98 as we do in 1.0

I dont have 0.98 around but i did compare 1.0 effects to bf2 ones aswell as creating some test effects. I tested it with the e_bhit_s_tarmac/ceramictiles by shooting at my feet as reported here.

Code: Select all

Effects                            Max  Min Size    #Textures Duration
PR 1.0                             120  60  5694kB      7       5
BF2                                120  100 110kB       5       <1
-1 big texture(short duration)     120  65  4294kB      6       5
-1 big texture(long duration)      120  95  4294kB      6       1
PR 1.0 Max quality: Medium         120  75  5694kB      7       5
Duration halved                    120  85  5694kB      7       2.5

Max/Min: maximum/minimum FPS, size: total size off all textures, #Textures: Amount of individual textures, Duration: total duration of effect
In this scenario the biggest fps drain was the duration of the effect and not the quality or amount of textures used. The new quality surely has an impact, especially for lower end PCs but does not seem to be the main factor.

[ballard_44]

First off, I did not test this on any of the fpstest's, but thought I'd share.

khamisiyah, coop 16, around 96 bots

Basically, I ran as a US soldier to the first cap flag and looked across the bridge and then didn't move.
FPS was in the low 50's.
Eventually, bots started surrounding me and the flags were capped.
The bots themselves lowered my FPS to low 40's.
Soon after they faced off at the bridge and a massive firefight ensued.
My FPS dropped to under 20 and stayed there.
Again I never moved.

2nd test was I gave bots nothing but stones to use. Nothing else was available to them.
Same FPS as above, until they started to fight each other.
When they used rocks, my FPS stayed in the low 40's.

Granted you'll never see this massive of a firefight in PR, but the sounds and effects certainly had an effect once they kicked into high gear.

Hope it helps!

BTW, watching bots use stones on each other and myself was quite amusing!

[Kingy]

Bots use a lot of processing power anyway, so having 96 bots on the field at a time is going to drop your fps regardless of the effects.

It's worth mentioning that the only way I could achieve any sort of playability on my old rig was by keeping everything on low (ofc) but I found that changing sounds from medium to low had a huge effect on my fps. Making it playable if rather dull.

[AfterDune]

Hmmm.... what happens when you set sounds to low? Less sounds at the same time or just less quality?

If the latter, I wonder what would happen if, for testing purposes, we would downgrade all sounds to something of lower quality and filesize.

[Spush]

One thing I would suggest is decreasing the draw calls. Every unique texture or object increases more load for the gpu. I'm pretty sure our draw calls are pretty bad at the moment.