Mac UT2K4 (Unreal Tournament 2004) demo BENCHMARK here

Discussion in 'Games' started by Santaduck, Feb 21, 2004.

  1. Santaduck macrumors 6502a

    Santaduck

    Joined:
    Oct 21, 2003
    Location:
    Honolulu
    #1
    I adapted the benchmark from Leonidas at 3dCenter.com for use on the Mac OS X version of the UT2004 demo.

    Includes a flyby benchmark and an assault benchmark, for resolutions ranging from 640x480 through 1600x1200.

    download it here:
    http://www.santaduck.com/

    I haven't fully tested it, and it isn't exactly user-friendly (you must use Terminal), but it works. Just follow the instructions and you should be ok.


    Frag Different

    _____

    UPDATE: I just found out that the benchmark doesn't work properly with the 3120.1 update.

    Specifically it doesn't exit UT2k4 immediately. If I can fix this, an update will be provided. For now the best & fastest way to exit without messing up the FPS too much is to hit tilde, then type EXIT (return) after the benchmark demo is finished playing.

    However, if you are going to report benchmark scores, please use the stock ut2k4demo 3120 (without 3120.1 update).
     
  2. Santaduck thread starter macrumors 6502a

    Santaduck

    Joined:
    Oct 21, 2003
    Location:
    Honolulu
    #2
    Here are initial results from my G4 Cube (stock & 1337), without 3120.1 update.

    Stock 500Mhz Cube G4/Stock cube Radeon/ 1G RAM/120G Maxtor (8M), Panther
    Flyby 800x600
    0.028127 / 4.847471 / 137.619797 fps -- Score = 4.844816
    Assault 800x600
    0.018739 / 2.102129 / 140.477615 fps -- Score = 2.101494


    PowerLogix 7457 1.4Ghz Cube G4/ATI Radeon 9000pro/1.5G RAM/120G Maxtor (8M), Panther
    Flyby 800x600
    0.060755 / 13.729904 / 591.040283 fps -- Score = 13.696004
    Assault 800x600
    0.054527 / 6.284725 / 590.607117 fps -- Score = 6.273344
     
  3. alphaone macrumors 6502

    Joined:
    Aug 12, 2003
    #3
    Wait...Do you have two Cubes or did you change the internals specifically for each benchmark:confused:
     
  4. Santaduck thread starter macrumors 6502a

    Santaduck

    Joined:
    Oct 21, 2003
    Location:
    Honolulu
    #4
    haha, my upgrades from OWC/macsales.com arrived yesterday in the mail. I delayed installing it until I got a benchmark on my stock setup. But I do have a stock setup cube at work as well with even lower specs (except it has a 23" cinema display).
     
  5. hvfsl macrumors 68000

    hvfsl

    Joined:
    Jul 9, 2001
    Location:
    London, UK
    #5
    If people want to compare with a PC (My PB G4 is not fast enough to run UT2004):

    UT2004 Build UT2004_Build_[2004-02-10_03.01]
    Windows 2000 5.0 (Build: 2195)
    AuthenticAMD PentiumPro-class processor @ 1841 MHz with 511MB RAM
    NVIDIA GeForce4 Ti 4200 with AGP8X (5655)

    Assault 1024x768 all settings max (flyby does not work)
    13.621860 / 32.278423 / 61.934376 fps rand[12750]
    Score = 32.320496
     
  6. Santaduck thread starter macrumors 6502a

    Santaduck

    Joined:
    Oct 21, 2003
    Location:
    Honolulu
    #6
    yeah there are some serious mac/Panther/G4/G5 or ATI optimization problems here.

    I'll run the PC demo on my 2.4G intel laptop at 800x600 and see what I get. Of course my cube is 3+ yrs old, and even w/the upgrade its memory is slower, and its logic board throughput is slower, and of course it's got a 32bit architecture. Still, that doesn't explain the vast performance hit on a mac. My UT2K3 benchmarks are in the 32FPS neighborhood, so something strange is up, esp. on that 21-player assault timedemo.
     
  7. Mav451 macrumors 68000

    Mav451

    Joined:
    Jul 1, 2003
    Location:
    Maryland
    #7
    I'm having trouble running the bench right...here are some of the strange results:

    (1)
    UT2004 Build UT2004_Build_[2004-02-10_03.01]
    Windows XP 5.1 (Build: 2600)
    AuthenticAMD PentiumPro-class processor @ 2224 MHz with 511MB RAM
    RADEON 8500 SERIES (6396)

    -benchmark -ini=UT2004.ini -1024x768 -seconds=10 -exec=demo_assault.txt

    127.494179 / 201.809692 / 395.506042 fps rand[11120]
    Score = 202.942139

    (2)
    UT2004 Build UT2004_Build_[2004-02-10_03.01]
    Windows XP 5.1 (Build: 2600)
    AuthenticAMD PentiumPro-class processor @ 2235 MHz with 511MB RAM
    RADEON 8500 SERIES (6396)

    -benchmark -ini=UT2004.ini -1024x768 -seconds=10 -exec=demo_flyby.txt

    159.415695 / 1757.576050 / 1.#INF00 fps rand[22491]
    Score = 1601.707153

    (3)
    UT2004 Build UT2004_Build_[2004-02-10_03.01]
    Windows XP 5.1 (Build: 2600)
    AuthenticAMD PentiumPro-class processor @ 2224 MHz with 511MB RAM
    RADEON 8500 SERIES (6396)

    -benchmark -ini=UT2004.ini -1024x768 -seconds=10 -exec=demo_flyby.txt

    118.961067 / 202.372986 / 431.202057 fps rand[11120]
    Score = 202.985107

    (4)
    UT2004 Build UT2004_Build_[2004-02-10_03.01]
    Windows XP 5.1 (Build: 2600)
    AuthenticAMD PentiumPro-class processor @ 2224 MHz with 511MB RAM
    RADEON 8500 SERIES (6396)

    -benchmark -ini=UT2004.ini -1024x768 -seconds=10 -exec=demo_assault.txt

    65.024971 / 1559.140259 / 1.#INF00 fps rand[26851]
    Score = 1336.129272

    *any help out there? Something is definitely up...i mean wut the heck is #inf mean??
     
  8. Santaduck thread starter macrumors 6502a

    Santaduck

    Joined:
    Oct 21, 2003
    Location:
    Honolulu
    #8
    UPDATE:

    I just made a new botmatch benchmark (CTF) that is compatible with Mac ut2k4's v3120.1 update! So go download it now =)

    I'm trying to get xlr8yourmac.com to list it at their benchmark site here.

    I got around the demorec playback problem because my CTF benchmark is designed like the ut2k3 ones, i.e. it is NOT playing a demorec. Again, it's available at santaduck.com. The download still includes the older benchmarks (assault & "flyby"). I will convert the CTF benchmark for use on windows today, and it will be available at the website as a separate download.



    MAV451 & your windows benchmark problem:
    lol this is a mac forum, sorry can't help you really with the windows version too much. :D I can only say the *inf usually means a VERY high number, which probably results because the 3dcenter (leonidas) benchmarks (i.e. assault & flyby) are demorec playbacks: this means you see a map load entry screen, all the while your FPS is being measured. In some configurations, your FPS may be super high, since all it is doing is presenting a static graphic.

    I'll convert my CTF benchmark for windows today, and that will avoid this INF problem. Keep checking at santaduck.com.
     
  9. Santaduck thread starter macrumors 6502a

    Santaduck

    Joined:
    Oct 21, 2003
    Location:
    Honolulu
    #9
    UPDATE: (sorry for all the updates, this should be the final version as I have solved all the problems and made it user-friendly).

    I finished coding a GUI botmatch benchmark for UT2004 that controls standard video settings, the "CTF Standard Benchmark". Download it at http://www.santaduck.com .

    How good can it get? Just point & click. No mucking around in Terminal Commandline. Also includes automatic installer built into the application, so no mucking around in System Packages either! Uses standard custom INI files for consistent video settings across all users from 640x480 through 1600x1200. A windows command version will be available here soon for comparisons.

    Notes: The CTF map was selected because it resulted in the most active action out of any of the maps. Most of the other maps were very large and task-oriented, which means the benchmark would just watch a bot sitting by itself shooting a gun installation, or idly standing still building something, or worse and more frequently just running alone across a large map. I could make other benchmarks, but realistically they would just suck as a measurement of system performance. The CTF map with 14 bots resulted in the most action at different spectator magnifications.

    This is why I have chosen NOT to include botmatch benchmarks of the other maps: this is the best map for a benchmark, and 14 botcount resulted in the best action.

    I feel that too many options make benchmarks useless for comparisons,so that's why there are only five, one for each resolution: 640x480, 800x600, 1024x768, 1280x960, and 1600x1200. Simple is the best.

    Unlike other UT benchmarks, you do not have to worry about your saved game settings because this benchmark uses customized INI files. This means that the results of the benchmark on any computer will be automatically set to the EXACT SAME video settings every single time on all computers. This eliminates the possibility that users will run the benchmarks using different video settings: this is a "Standardized" benchmark, whose results can be compared meaningfully.

    (These default settings are: color depth 32bit, texture detail Normal, Character Detail Normal, World Detail High, Physics Detail Normal, Dynamic Mesh LOD Normal, Decal Stay Normal, Character Shadows Full, Decals (X), Dynamic Lighting (X), Cornoas (X), Trilinear Filtering (X), Projectors (X), Foliage (X), Weather Effects (X), Fog Distance Bar all the way to the right, Low Sound Detail ( ).)
    regards,
    _________
     

Share This Page