Performance Test II - pr_fpstest2

Post by **AfterDune** » 2013-11-07 16:00

Following up on my previous attempt, here's another attempt at increasing PR's performance.

So, what is different?

All files again will be loaded from your PR installation. But this time, we're using the launcher again, as performance was exactly the same while using the old pr.exe.

This test only features five maps:
- Black Gold
- Dovre,
- Dragon Fly
- Fallujah West
- Khamisiyah

I can add more maps on demand, but I'd like to first try out the ones above.

All static (=placed by the mapper) ambient effects like fire and smoke have been removed. Waterfalls or fountains are still there.

If a map crashes on you, try to load another one. They should all work, but for some reason some people crash now and then

.

Download and "installation"

Download this file: https://dl.dropboxusercontent.com/u/104 ... stest2.rar (8 MB)

It contains a folder "pr_fpstest2". 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_fpstest2" using PRLauncher.exe, located here:
32-bit: bf2-dir/mods/pr_fpstest2/bin/bin32/PRLauncher.exe
64-bit: bf2-dir/mods/pr_fpstest2/bin/bin64/PRLauncher.exe

Dedicated server

I'm pretty sure this testmod also works on a dedicated server. If someone is willing to host "pr_fpstest2" on a dedicated server, feel free to do so.

So if you see a dedicated server running, please join it.

What to test?

1.
Run a map in normal PR 1.0 and see what the FPS is like. You can check the FPS by opening your console (default is ~ or END-key). Then type: renderer.drawfps 1
This will display a little red text on the topleft of your screen. This is enabled by default in pr_fpstest however.
Anyway, check if there are differences in FPS when running both mods.

2.
Loading times, are they any different? I don't think so, but always good to compare some numbers.

3.
Play a COOP game in PR 1.0 and then pr_fpstest and see what it does to performance.

4.
If someone runs a dedicated server, join it and play some games.

If you've tried all or some of the above, please post back here and share your findings.

Prevtzer · Post by **Prevtzer** » 2013-11-07 16:42

Fallujah West- 48 bots
Looking away from the city at the US spawn point- 65fps, looking towards the city- 15fps.
Same for fpstest2, also no noticeable change in loading times.

Post by **KaB** » 2013-11-07 22:19

Not much different. The weird thing of this is that I can get an aweful framerate in empty areas while I get decent fps in areas full of buildings and objects.

Can't you guys code a software that would make the players rotating constantly in all points of the map ? In other words having the number of fps of all the possible views of the map (each point would register the average of the fps of every angles).
With those numbers you'd make a sort of graph incoprated into the map where the highest number goes green and the lowest goes red. Then you could easily see what areas are the more affected.

Post by **Rusty_42** » 2013-11-08 01:48

No difference on local. As I know, most problems with FPS starts when we have stack of players in one place. So we need somehow simulate this on dedicated server and run tests.

Post by **Mats391** » 2013-11-08 20:19

Tested it on 2 maps:

Code: Select all

Fallujah INS inf- Top of minaret
Normal:		Min: 31 - Max: 63
FPS_TEST:	Min: 35 - Max: 71

Black Gold INS - F7kp1 on tower
Normal:		Min: 58 - Max: 96
FPS_TEST:	Min: 54 - Max: 126

The impact of the static effects does not seem too big. On Black Gold the minimum was reached when facing the city, maximum when looking towards the desert/bombed tanks. Since there are some smoke clouds around those the maximum on fps_test was obviously higher. Same for Fallujah.

Post by **K4on** » 2013-11-09 08:50

There were quite a few players complaining about the bad FPS, but most of them haven't even made the effort to give feedback here or in the other fps test thread. Com'on guys, help us - thanks to the others who did already!

PS - This is quite a good example of how to post feedback properly:

Mats391 wrote:Tested it on 2 maps:

Code: Select all

Fallujah INS inf- Top of minaret
Normal:		Min: 31 - Max: 63
FPS_TEST:	Min: 35 - Max: 71

Black Gold INS - F7kp1 on tower
Normal:		Min: 58 - Max: 96
FPS_TEST:	Min: 54 - Max: 126

Souls Of Mischief · Post by **Souls Of Mischief** » 2013-11-09 11:32

Load times (Fallujah, 48 bots)

FPS Test 2 - around 2:10 Normal 2:02

FPS (Fallujah 48, bots)

Facing SW from the ammo resupply station.

FPS Test 2 - High 80, dipped down to 48 (heard a lot of gunfire in the city)
Normal - High 75, dipped down to 51 (same)

Facing the city (NE) from the confines of the base.

FPS Test 2 - lowest 21, highest 26.7
Normal - 23-24.1, dipped down all the way to 15.1 (explosions, gunfire, heard one vehicle exploding in the city)

Code: Select all

System Information
------------------
  Operating System: Microsoft Windows 7 Home Premium  (6.1.7601)
      Architecture: 64-bit
          Language: *******
       Motherboard: Dell Inc. 0FXK2Y
         Processor: Intel(R) Core(TM) i5-2450M CPU @ 2.50GHz (Physical: 2, Logical: 4)
            Memory: 6,00 GB
      DIMM Modules: DIMM_A: 2,00 GB @ 1333 MHz
                    DIMM_B: 4,00 GB @ 1333 MHz
         Page File: 9,00 GB
    .NET Framework: 4.0

Display Information
-------------------
 Display Device(s): Generic PnP Monitor on Intel(R) HD Graphics Family
   Display Mode(s): 1366 x 768 (32 bit) @ 60 Hz
    Driver Version: 8.15.10.2361
    Display Memory: 2,77 GB
     Multisampling: 2, 4, 8
               DPI: 96 (100%)

Audio Information
-----------------
  Primary Playback: Speakers / Headphones (IDT High
 Primary Recording: Microphone Array (IDT High Defi
               EAX: True
           EAX 1.0: False
           EAX 2.0: True
           EAX 3.0: False
           EAX 4.0: False
           EAX 5.0: False
             X-RAM: False

Disk Information
----------------
      Install Path: C:\Games\Battlefield 2
                    Free: 42,56 GB, Total: 683,89 GB, NTFS
          Mod Path: C:\Games\Battlefield 2\mods\pr
                    Free: 42,56 GB, Total: 683,89 GB, NTFS
     Profiles Path: C:\Users\Niklāvs\Documents\ProjectReality\Profiles
                    Free: 42,56 GB, Total: 683,89 GB, NTFS
    Update DL Path: C:\Users\Niklāvs\AppData\Local\Project Reality\Project Reality BF2\Downloads
                    Free: 42,56 GB, Total: 683,89 GB, NTFS
   Update Log Path: C:\Users\Niklāvs\AppData\Local\Project Reality\Project Reality BF2\UpdateLogs
                    Free: 42,56 GB, Total: 683,89 GB, NTFS

Game Information
----------------
      CD Key Valid: True
    Installed Mods: bf2, fh2, pr, pr_fpstest2, xpack
       Current Mod: pr
       BF2 Version: BF2 1.5
        PR Version: 1.0.36.0
          Language: english
   Debug Available: False
     Large Address: True
Supported Hardware: None

Profile Information
-------------------
      Profile 0001: *********
              Type: Online
 Last Used Profile: True
        View Intro: False
        Fullscreen: True
      Display Mode: 1366x768@60Hz
Display Mode Valid: True
   Graphics Scheme: Custom
     Multisampling: 4x
             VSync: False
   Terrain Quality: Low
   Effects Quality: High
  Geometry Quality: Low
   Texture Quality: Low
  Lighting Quality: Low
   Dynamic Shadows: Off
    Dynamic Lights: Off
 Texture Filtering: Medium
    Audio Provider: Software
    Provider Valid: True
     Audio Quality: Medium
               EAX: True

Post by **ballard_44** » 2013-11-09 12:10

FYI for everybody,

Bots can be resource heavy.
More bots and higher difficulty can grind even the better computers to a halt if certain scenarios play out.

Might not be the best way to help these tests???

Prevtzer · Post by **Prevtzer** » 2013-11-09 16:21

[R-DEV]K4on wrote:There were quite a few players complaining about the bad FPS, but most of them haven't even made the effort to give feedback here or in the other fps test thread. Com'on guys, help us - thanks to the others who did already!

The problem is it's been a while since 1.0 was released so 90% of people that suddenly couldn't play already left.

Post by **K4on** » 2013-11-09 16:32

ballard_44 wrote:Bots can be resource heavy.
More bots and higher difficulty can grind even the better computers to a halt if certain scenarios play out.

Thats is true of course, but a big deal of the performance could be about effects. That is what we want to test in this version. If you have no fights going on, there are no explosion effects etc.
A few bots, like 20 or so should be easy to handle, even for older cpus.

Apart, please try to run maps like Khamisiyah and DragonFly and tell us about the performance there. Fallujah was always laggy due to its statics/objects, but this time we want to have feedback on the effects. So try the other maps aswell please.

Souls Of Mischief · Post by **Souls Of Mischief** » 2013-11-09 16:54

'[R-DEV wrote:K4on;1962522']Thats is true of course, but a big deal of the performance could be about effects. That is what we want to test in this version. If you have no fights going on, there are no explosion effects etc.
A few bots, like 20 or so should be easy to handle, even for older cpus.

Apart, please try to run maps like Khamisiyah and DragonFly and tell us about the performance there. Fallujah was always laggy due to its statics/objects, but this time we want to have feedback on the effects. So try the other maps aswell please.

Wasn't for me in 0.98 on medium/high.

Post by **Mats391** » 2013-11-09 19:00

[R-DEV]K4on wrote:Thats is true of course, but a big deal of the performance could be about effects. That is what we want to test in this version. If you have no fights going on, there are no explosion effects etc.
A few bots, like 20 or so should be easy to handle, even for older cpus.

Apart, please try to run maps like Khamisiyah and DragonFly and tell us about the performance there. Fallujah was always laggy due to its statics/objects, but this time we want to have feedback on the effects. So try the other maps aswell please.

I tested Khamisiyah and got same results as on block gold:

Code: Select all

Khamisiyah AAS - I8kp9 on watertower
Normal:		Min: 70 - Max: 130
FPS_TEST:	Min: 77 - Max: 150

The minimum was when looking towards lots of statics (bunkers+chem factory). Obviously the static ambient effects have an impact on the performance, but it doesnt seem to be a huge fps killer for me as the minimum values are similar.
Also i think this test should be done w/o bots. It is to test map performance and impact of ambient effects, bots only screw that up.

Heavy Death · Post by **Heavy Death** » 2013-11-09 19:22

Since Boris found out that supply depot's shadow render caused the drop, maybe try removing them on these maps youre dfiddling around with? Wouldn't hurt, eh.

Post by **Mats391** » 2013-11-09 19:24

Heavy Death wrote:Since Boris found out that supply depot's shadow render caused the drop, maybe try removing them on these maps youre dfiddling around with? Wouldn't hurt, eh.

That was related to a wrong SLI setup. For single card users the impact of shadows isnt that big.

Boris · Post by **Boris** » 2013-11-11 13:22

A bit late to the party here, but here you go:

Code: Select all

ver		start	map		load	fps	gpu	notes
fpstest1	24s	fallujah	39s	94	61%
fpstest2	22s	fallujah	39s	96	61%	bin32
1036		22s	fallujah	38s	96	61%
0981		15s	fallujah	29s	107	68%

ver = pr version
start = initial load time to get to lobby
map = map used to test
load = map load time
fps = map fps reading taken at a specific location
gpu = gpu-z "gpu load" reading
notes = additional info

Couldn't run the fpstest2 bin64 at the time, PRLauncher failing at start-up with error:

Code: Select all

Project Reality Log File
PRLauncher.exe started at: 2013-11-11T12:29:13

Error: 6: Unable to find localization. Please reinstall Project Reality: BF2

I guess there's a registry entry missing somewhere or something (HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Nls\CustomLocale\?).

System:

Code: Select all

System Information
------------------
  Operating System: Microsoft Windows 7 Ultimate  (6.1.7601)
      Architecture: 64-bit
          Language: English (United Kingdom)
       Motherboard: ASUSTeK COMPUTER INC. MAXIMUS V EXTREME
         Processor: Intel(R) Core(TM) i5-3570K CPU @ 3.40GHz (Physical: 4, Logical: 4)
            Memory: 8.00 GB
      DIMM Modules: ChannelA-DIMM1: 4.00 GB @ 1867 MHz
                    ChannelB-DIMM1: 4.00 GB @ 1867 MHz
         Page File: -1908736.00 B
    .NET Framework: 4.0

Display Information
-------------------
 Display Device(s): Generic PnP Monitor on NVIDIA GeForce GTS 250
   Display Mode(s): 1280 x 1024 (32 bit) @ 60 Hz
    Driver Version: 9.18.13.2723 (327.23)
    Display Memory: 512.00 MB
     Multisampling: 2, 4, 8
               DPI: 96 (100%)

Audio Information
-----------------
  Primary Playback: Speakers (Realtek High Definiti
 Primary Recording: Microphone (Realtek High Defini
               EAX: True
           EAX 1.0: False
           EAX 2.0: True
           EAX 3.0: False
           EAX 4.0: False
           EAX 5.0: False
             X-RAM: False

Disk Information
----------------
      Install Path: D:\games\bf2pr
                    Free: 25.19 GB, Total: 140.00 GB, NTFS
          Mod Path: D:\games\bf2pr\mods\pr
                    Free: 25.19 GB, Total: 140.00 GB, NTFS
     Profiles Path: C:\Users\User\Documents\ProjectReality\Profiles
                    Free: 3.41 GB, Total: 25.00 GB, NTFS
    Update DL Path: D:\games\bf2pr\binaries
                    Free: 25.19 GB, Total: 140.00 GB, NTFS
   Update Log Path: D:\games\bf2pr\logs
                    Free: 25.19 GB, Total: 140.00 GB, NTFS

Game Information
----------------
      CD Key Valid: True
    Installed Mods: bf2, pr, pr_fpstest, pr_fpstest2, xpack
       Current Mod: pr
       BF2 Version: BF2 1.5
        PR Version: 1.0.36.0
          Language: english
   Debug Available: True
     Large Address: True
Supported Hardware: None

Profile Information
-------------------
      Profile 0001: defaultPlayer
              Type: Offline
 Last Used Profile: True
        View Intro: False
        Fullscreen: True
      Display Mode: 1280x1024@60Hz
Display Mode Valid: True
   Graphics Scheme: High
     Multisampling: 4x
             VSync: False
   Terrain Quality: High
   Effects Quality: High
  Geometry Quality: High
   Texture Quality: High
  Lighting Quality: High
   Dynamic Shadows: High
    Dynamic Lights: High
 Texture Filtering: High
    Audio Provider: Hardware
    Provider Valid: True
     Audio Quality: High
               EAX: True

Post by **LITOralis.nMd** » 2013-11-11 16:33

Boris,

there is an opensource project that is a addon to FRAPS,
FRAFS Bench Viewer | Free System Administration software downloads at SourceForge.net

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

I have a question to long time BF2 players, can a Battlerecorder scene recorded in PR1.0.36 be transferred to the mods/pr_test directory and then replayed?
If that is possible, creating several battlerecorder scenes with proper gameplay action, then transferring it between subdirectory of the BF2\mods\ directory, could speed up testing abilities going forward.

I am in no way capable of even using battlerecorder properly, but if the program is mod subdirectory agnostic, looks like we could in fact create benchmark demo scenes and transfer the scenes between test builds.

Post by **LITOralis.nMd** » 2013-11-11 17:23

Using the recording techniques from the FCAT instructions:
http://international.download.nvidia.co ... _Guide.pdf

But changing out FCAT for FRAFS, since none of us have the US$3500 needed to make a FCAT recording PC,

we can use some of the same techniques
specifically
FRAPS
VirtualDub
Excel
FRAFS

record the video with ShadowPlay,
bind the shadowplay keybind and the FRAPs benchmark keybind to same key
OR
if someone has a proper twitch.tv streaming PC with a proper capture card, you can work like that.

And Frafs intepretation of FRAPS benchmark Frametimes data

Frafs:
https://frapsforum.com/threads/new-tool ... iles.2121/
FRAFS Bench Viewer | Free System Administration software downloads at SourceForge.net

...

Get that all combined and working, and we should be able to get a better grasp of frame buffer output.

Should then be able to find the exact effect/texture/lod/model/ etc that causes frame rate drops.

---

I'm at work and spent 30 minutes reading about gpu render scan out T_render time. sorry if only semi-coherent.

Post by **LITOralis.nMd** » 2013-11-11 17:26

Ramiel Co-Op Alt Inf, 32 bots, local
between control sector and Objective Alpha

Notice the cliffs in minimum FPS, those are the locations we can work on isolating to find the exact problems.

Boris · Post by **Boris** » 2013-11-11 17:29

Considered running a bf2demo here, but didn't look too hard into making it work. I think at least one problem to overcome is that there is no fixed 1st-person 'in-eyes' view for playback (as far as I know), so it might not be so simple to recreate the same conditions for each run. I would certainly use this to test though if it were possible to work around these issues.

timedemo 1
map demo1.dm2

*keep dreaming*

Post by **LITOralis.nMd** » 2013-11-11 19:58

I think there is a way to create a fly by wire script for the camera view in battlerecorder.
Yes, it is third person view, but it should be a fixed path that is exactly the same on each replay.

o If working from a client that has downloaded a demo from the game’s front end:
o Create a "Demos" folder in the root of the game install folder.
o Move the demos you wish to render from into the demos folder from their normal location in you’re my Documents folder.
o Drop the demo.cmd folder found below into the root of the game install folder.
Launch the demo recording as follows o Launch the Command Prompt and browse to the game’s install directory. o E.g. type "cd c:\Program Files\EA Games\Battlefield 2" and hit ENTER
Type "demo.cmd +c " and press enter. o E.g. "demo.cmd +c mydemo mydemocam" will record a Demos\mydemocam.bf2cam for demo Demos\mydemo.bf2demo
o Before quitting the game, wait for the message 'camera file closed' in the console or the camera file will be broken.
Playing back a demo with pre-recorded camera movements
This mode combines the previous steps and shows you a take with both the demo and camera tracks active. o Launch the game as follows: demo.cmd o To watch the demo with the default recorded camera use: demo.cmd -def
Rendering images to disk
This mode writes out an AVI file to disk based on a demo file and a recorded camera file. o Launch the game as follows: demo.cmd +r o A dialog box pops up where you can select encoder (Xvid or Divx is recommended) and the quality settings for that encoder.
If you want to stop rendering before the end of the demo, press 'R'

" Create a "Demos" folder in the root of the game install folder. " I don't know where that should be in PR 1.0 , need help from someone who knows how the directory changes in PR 1.0 affect usage of battlerecorder.

Performance Test II - pr_fpstest2

Performance Test II - pr_fpstest2

Re: Performance Test II - pr_fpstest2

Re: Performance Test II - pr_fpstest2

Re: Performance Test II - pr_fpstest2

Re: Performance Test II - pr_fpstest2

Re: Performance Test II - pr_fpstest2

Re: Performance Test II - pr_fpstest2

Re: Performance Test II - pr_fpstest2

Re: Performance Test II - pr_fpstest2

Re: Performance Test II - pr_fpstest2

Re: Performance Test II - pr_fpstest2

Re: Performance Test II - pr_fpstest2

Re: Performance Test II - pr_fpstest2

Re: Performance Test II - pr_fpstest2

Re: Performance Test II - pr_fpstest2

Re: Performance Test II - pr_fpstest2

Re: Performance Test II - pr_fpstest2

Re: Performance Test II - pr_fpstest2

Re: Performance Test II - pr_fpstest2

Re: Performance Test II - pr_fpstest2