MayaCarBench: real world performance in Autodesk Maya

Discussion in 'Benchmark Mayhem' started by NormanBates, Oct 24, 2009.

  1. n4d444

    n4d444 Active Member

    Messages:
    71
    Likes Received:
    0
    GPU:
    8800GT@FX3700
    Big companies ("big" managers) think big. They think a few new stupid features are gonna make them more customers/profit than optimizing their existing program. All programs that have been bought by a big company suffer like this.
    I didn't test the scene on my computers but from what I can see looking at these benches is that it's a bad scene. It not so much a bad scene as much as Maya lack optimizations and your CPU and HDD make a larger impact on your frame rate than your graphic card does.
    Trust me in a regular scene (where you aren't bottlenecked by CPU/HDD/lack of optimization) a real/modded quadro (no difference in Maya from what I have seen untill now) is at least 2-3 times faster than a regular geforce or a real/modded firegl.
     
  2. NormanBates

    NormanBates Master Guru

    Messages:
    376
    Likes Received:
    0
    GPU:
    ATI HD3870 / ATI HD4670
    well, this scene is not far from what I would be using at those times when I get an urge to scream at my screen (I'm not interested in having more fps when I already have *enough* fps)

    what I can picture quite easily is that, as I'm not a real pro, I'm mostly self-taught and still in my learning process, what I work with may be completely different from what a pro user would be working with

    maybe you would be generous enough to provide us with a "regular scene" which is hard enough on any system to be a useful benchmark?
     
  3. Maryus3D

    Maryus3D Active Member

    Messages:
    97
    Likes Received:
    0
    GPU:
    Zotac GTX 285
    Remember the bike scene? The textured one? :D I guess we should use it as a benchmark ... on a regular card the scene has around 7-8 fps but on Quadro mod has over 20fps :D.
     
  4. n4d444

    n4d444 Active Member

    Messages:
    71
    Likes Received:
    0
    GPU:
    8800GT@FX3700
    These pictures just show how badly optimized Maya is.
    You all guys should also bench in wireframe. That seams to show a cards power a bit.

    [​IMG]
    [​IMG]

    Strange how 3ds max is so much faster in shaded view than in wirframe. Maya shaded view performance is horrible.
    I just started learning Maya (I am just curious and mby I'll need it one day). I am a 3ds max user for many years now. I was amazed how completely similar the two programs are, even some of the shortcuts are the same.

    edit: btw this is a modded quadro FX3700 (8800GT) not a real one (tested under WinXP x64 SP2 with Maya 2009 at 1680x1050 with Maya profile in nv control panel I also forgot to mention I have 8GB of ram)
    edit: To answer your question. It's about CPU power. Quadros are all gonna performe similary here cos of Maya bad optimization. If I put an FX1700 in my comp I would probably get exactly the same results in shaded view but a bit smaller results in wireframe. I am gonna prove this some other time.
     
    Last edited: Oct 28, 2009

  5. NormanBates

    NormanBates Master Guru

    Messages:
    376
    Likes Received:
    0
    GPU:
    ATI HD3870 / ATI HD4670
    yes, I remember the maya scene, and specially the textured one; in fact, I textured it, in about 5 minutes: I googled something like "brass", got a couple of really big bitmaps, created some new materials with them, and applied them mostly at random, with no meaningful UV mapping at all :)

    so it's not a bad benchmark, but I felt it had some potential to be misleading

    this time the models are realistically textured: someone (not me) has taken his/her time to apply those materials correctly

    anyways, it may not be a bad idea to collect all of the BikeBench results and create a table like the one I posted at the beginning of this thread, to see if we can make some sense out of them; I'll give it a try...
     
  6. NormanBates

    NormanBates Master Guru

    Messages:
    376
    Likes Received:
    0
    GPU:
    ATI HD3870 / ATI HD4670
    well, I don't work on wireframe that often, and it's not usually there where I'm craving for a bit more performance, so, if it doesn't show a big difference because of the GPU, that's the result I'm taking home

    two things seem clear now:
    * CPU is really important: all nvidia results are ordered by CPU speed; ati (me) comes behind, far, far away
    * it seems quite probable that I messed up the texture part of the scene: given the huge difference between shaded and wireframe, it's quite probable that textures are too big even for a GPU with 1GB of memory

    in any case, thanks for your results, really interesting: you're beating the real Quadros!

    plus: could you or Maryus3D test again without the mod? we're missing some bare modded/unmodded comparisons for the nvidia cards (though that 9500 result seems to imply that the mod is nearly irrelevant in this scenario)
     
  7. n4d444

    n4d444 Active Member

    Messages:
    71
    Likes Received:
    0
    GPU:
    8800GT@FX3700
    This picture reveals many things. What can I say? I just love Photoshop.:nerd:
    Everyone should make ALL the same benchmarks as I did. This way paints a better overall picture. I just deleted all the jpg/bmp/tga/png/tif... in the folders and loaded the scene for the "DELETED TEXTURES" benchmark.

    [​IMG]

    edit: This is my HP xw4600 Workstation at work (nothing overclocked)
    edit2: I checked video memory usage with everest ultimate edition (OSD panel)
     
    Last edited: Oct 29, 2009
  8. NormanBates

    NormanBates Master Guru

    Messages:
    376
    Likes Received:
    0
    GPU:
    ATI HD3870 / ATI HD4670
    nice, I added those results to the initial table

    how did you check the VGA RAM status?
     
  9. NormanBates

    NormanBates Master Guru

    Messages:
    376
    Likes Received:
    0
    GPU:
    ATI HD3870 / ATI HD4670
  10. Maryus3D

    Maryus3D Active Member

    Messages:
    97
    Likes Received:
    0
    GPU:
    Zotac GTX 285
    Last edited: Oct 29, 2009

  11. wpgscorpion

    wpgscorpion New Member

    Messages:
    1
    Likes Received:
    0
    GPU:
    eVga GeForce 9800 GT
    With unmodded drivers, and cpu and video card not overclocked.
    System is newly built i7 860, Gigabyte P55M-UD2, eVga GeForce 9800GT 512MB, 4 x 2GB Patriot Viper2 DDR3 1600Mhz XMP ram, Win 7 Professional 64 bit.

    I installed the Maya 2010 evalution version and saw a steady 6.8 fps on the cars\airplane bench.
     
    Last edited: Oct 31, 2009
  12. NormanBates

    NormanBates Master Guru

    Messages:
    376
    Likes Received:
    0
    GPU:
    ATI HD3870 / ATI HD4670
    one of my systems finished its work, so I run a bunch of tests:

    * my main result ------------- 4.4 fps
    * slow GPU down by 50% (both chip and its memory) -- 4.4fps
    * slow system memory down -- 4.3 fps
    * clock CPU back at 2.4 GHz -- 2.7 fps

    (they are non-cumulative, i.e., when I clocked the memory down the GPU was running back at its usual speed)

    conclusion: in such a scene, your CPU is the bottleneck; BUT still nvidia will do much better than ati

    edit: thanks wpgscorpion, I add your results to the table too
     
  13. kocha

    kocha Member

    Messages:
    12
    Likes Received:
    0
    GPU:
    NVidia Quadro FX4600
    Maybe little offtopic....

    Did you try SPECapc for Maya. Not SPECviewperf, but SPECapc...

    Main difference between theese two is that SPECviewperf "emulates" Maya "engine", and SPECapc use real Maya installation to run the test. It works on 2010 version too. Benchmark is free to download...

    Code:
    http://www.spec.org/gwpg/apc.static/maya2009info.html
     
  14. NormanBates

    NormanBates Master Guru

    Messages:
    376
    Likes Received:
    0
    GPU:
    ATI HD3870 / ATI HD4670
    no, I haven't tried it, because:
    * it requires maya 2009 (I'm still on 2008)
    * I doubt it's useful, it has some new scenes but it also keeps some old ones (hands, ants), which are not really useful




    edit: by the way, I got everest, and couldn't find out how to read the GPURAM usage; maybe that's because it's a trial version, some indicators only show "TRIAL VERSION"
     
    Last edited: Oct 30, 2009
  15. CADCAT

    CADCAT Guest

    Messages:
    232
    Likes Received:
    0
    GPU:
    Mobility HD5870 1GB GDDR5
    Good evening,

    here are the results for HD4870 and FireGL V7700:

    HD4870 ; Catalyst 9.10 ; 6.0-6.1 FPS ; GPU Usage: ~25%
    HD4870 ; FirePro 8.633 GellmaR ; 6.0 FPS ; GPU Usage: ~25%

    FireGL V7700 ; FirePro 8.633 ; 6.0 FPS ; GPU Usage: ~ 23% (?)

    HD3870 ; Catalyst 9.10 ; ??
    HD3870 ; FirePro 8.633 GellmaR ; ??
     

  16. NormanBates

    NormanBates Master Guru

    Messages:
    376
    Likes Received:
    0
    GPU:
    ATI HD3870 / ATI HD4670
    great results, cadcat: they pretty much paint the whole ATI picture on this side

    the conclusion is final: I'll paste it on post #1
     
  17. NormanBates

    NormanBates Master Guru

    Messages:
    376
    Likes Received:
    0
    GPU:
    ATI HD3870 / ATI HD4670
  18. Maryus3D

    Maryus3D Active Member

    Messages:
    97
    Likes Received:
    0
    GPU:
    Zotac GTX 285
    I have 7 fps, Windows 7 x64 182.08 patched Q6600@2,4 and 1680x1050.
     
  19. NormanBates

    NormanBates Master Guru

    Messages:
    376
    Likes Received:
    0
    GPU:
    ATI HD3870 / ATI HD4670
    I updated the results table

    I guess they patched something, nice to see w7 catching up
     
  20. Beavis

    Beavis Ancient Guru

    Messages:
    2,029
    Likes Received:
    1
    GPU:
    Quadro K620
    Awesome thread. I'm looking out for a new GFX card so this will help me out immensly! Thanks!
     

Share This Page