Maya 6 - 7900GT or 6800 Ultra -> QuadroFX 4000

Discussion in 'RivaTuner Advanced Discussion forum' started by augie, Aug 11, 2006.

  1. augie

    augie Active Member

    Messages:
    58
    Likes Received:
    0
    GPU:
    Geforce 7900GT @ 630MHz
    Wow, those are some serious numbers! Even FX4500 didn't reach as high a score as yours in this article:
    http://features.cgsociety.org/story_custom.php?story_id=3321&page=3
    System was a Dual Opteron 275 (2.2GHz, 1MB L2 cache, dual core)

    Those maya bench scores are literally 12X mine in Maya... Obviously, you have faster CPU as well.

    The link you provided was very helpful! :) The benchmarks there were using a 7900GT card, and even on a Conroe x6800 they only go up to 24.160 in maya-02... 122.7!

    Okay, that's DEFINITELY some difference between Geforce and Quadro (in SPECViewPerf 9 at least.) Up in the CGSociety article I linked, I also see that trend of Quadro > GF, but when they ran the Maya 6.5 benchmark (seperate SPEC.org benchmark), GF performed just as well as Quadro.

    Thanks so much for posting this. Now I think I have a clearer picture what I'm going to have to do to answer this question completely. I tracked down a used 6800 Ultra and ordered it. I am going to SoftQuadro it up to a QFX4000 and do the SpecViewPerf benchmarks and the SPEC Maya benchmark in Quadro and Geforce mode. Then if I can find somewhere with a decent return policy (< 15% restocking fee), buy a Geforce 7900GT and perform the same two benchmarks.

    Augie
     
  2. Stefan_VTi

    Stefan_VTi Member

    Messages:
    47
    Likes Received:
    0
    GPU:
    PNY Quadro FX 3500
    A quick link to my custom performance profile (used for these tests):

    http://www.engit.nl/quadro/nvwsapps.zip

    1. Install Quadro driver (preferably 91.36)
    2. Extract the file to your system32 folder.
    3. Open your quadro properties & pick the 'perfomance - single display' profile (do not have any 3d app open!)
     
  3. Stefan_VTi

    Stefan_VTi Member

    Messages:
    47
    Likes Received:
    0
    GPU:
    PNY Quadro FX 3500
    I checked the SPEC site, because indeed those FX 4500 results you posted looked so low, but as you can see here the FX 3500 is king in maya. The key seems to be the 90 series driver.
     
  4. augie

    augie Active Member

    Messages:
    58
    Likes Received:
    0
    GPU:
    Geforce 7900GT @ 630MHz
    Indeed, there's some serious differences between those systems...
    IBM IntelliStation Z Pro 9228 3.0GHz NVIDIA Quadro FX 3500 = 106.91
    Dell Precision Workstation 380 3.80GHz / nVIDIA Quadro FX 3500 = 29.45
    HP xw6400 Workstation 3.0GHz nVIDIA Quadro FX 4500 = 29.61

    All three use a single dual-core Xeon. The Dell has a 3.8GHz vs the other two with 3.0GHz procs, yet those numbers are insanely different. And yes, the top scoring 2 systems are using a 91.28 driver.

    FX4500 is based on NV70 core (7800 line of cards)
    FX3500 is based on NV71 core (7900 line of cards) and is clocked slower

    Other benches I've seen with these two cards shows the 4500 on top most of the time, but not by much.
    http://www.xbitlabs.com/articles/video/print/quadrofx-vs-firegl.html

    I would assume the 4500 would get 100+ SPEC9 scores too with the 91.28 driver?

    All that said, this makes the situation even more confusing! :O
    Did you notice the SPEC9 scores you linked for the Quadro workstations using older drivers were not much faster than the Geforce 7900GT scores you posted from GamePyre?

    I wonder what driver they were using... Hard to believe the 91.28 driver really be 5-6X as fast as 8x.xx in Maya benches.
     

  5. upf

    upf New Member

    Messages:
    7
    Likes Received:
    0
    GPU:
    6800GT > Quadro 4000
    Hi guys, I have a 6800GT PCI-E (NV40 ,A1) softmoded to QUADRO 4000 and this is the results of SPECopc 9 test, before and after :

    before
    maya-02 Weighted Geometric Mean = 17.89

    after
    maya-02 Weighted Geometric Mean = 32.50

    Driver version 91.36
    One thing - I did not notice the difference running the scenes (~2 mill poligons)
     
  6. augie

    augie Active Member

    Messages:
    58
    Likes Received:
    0
    GPU:
    Geforce 7900GT @ 630MHz
    55% performance boost in SPECViewPerf 9. That's pretty good. But as you say, not much difference in Maya scenes? What's your system config? Your ~FX4000 is getting a score about 1/4 (!) of Stefan_VTi's FX3500.

    The discrepancy between SPECViewPerf 9 results and actual Maya usage results seems to be quite large, judging from my observations on my 5900/FX3000, your 6800GT/FX4000, and even the CGSociety link I posted in an above message. (CGSociety used SPECapc for Maya 6.5 and all GF7800GTX & SLI came out equal to all FX4500,4400,1400,540,FireGL 7100)

    What version of Maya do you have? If you want to do a little more benching, maybe you could try SPECapc for Maya 6.5 @ http://www.spec.org/gpc/downloadindex.html

    They also have an older SPECapc for Maya 6 (if you don't have 6.5+) @ http://www.spec.org/gpc/apc.static/prev_benchmarks.html

    This is the closest thing to a real-world Maya usage benchmark I know of.
     
  7. upf

    upf New Member

    Messages:
    7
    Likes Received:
    0
    GPU:
    6800GT > Quadro 4000
    My sistem :

    Pentium D 820 2.8 Ghz
    Motherboard Intel Desktop Boards D945 GNT
    Graphics Card ASUS EN6800GT PCI_E
    Memory 2GB DDR2
    I use MAYA 6.0.1

    I'l download the SPECapc for Maya 6 and I'l post the test results.
     
    Last edited: Aug 16, 2006
  8. augie

    augie Active Member

    Messages:
    58
    Likes Received:
    0
    GPU:
    Geforce 7900GT @ 630MHz
    http://www.spec.org/gpc/opc.data/vp9/maya-perf.html
    Seems this page has been updated since yesterday with a new UBER high performing system - #1

    SPECViewPerf 9 - Maya-02 benchmarks:
    1. IBM IntelliStation A Pro 3.0 GHz NVIDIA Quadro FX 4500 X2 = 185.06 (91.31 driver) - (AMD Opteron Processor Model 256, 3.0GHz)
    2. HP Workstation xw9300 Opteron 256 3.0GHz Quadro FX 4500 x 2 = 24.08 (84.26 driver)

    3. IBM IntelliStation Z Pro 9228 3.0GHz NVIDIA Quadro FX 3500 = 106.91 (91.28 driver)
    4. Dell Precision Workstation 380 3.80GHz / nVIDIA Quadro FX 3500 = 29.45 (8x.xx driver)
    5. HP xw6400 Workstation 3.0GHz nVIDIA Quadro FX 4500 = 29.61 (8x.xx driver)

    #1 & #2 use identical Opterons, yet LOOK AT THE DIFFERENCE! Driver revision is the main difference here.
    All #2, #3, & #4 use a single dual-core Xeon. The Dell has a 3.8GHz vs the other two with 3.0GHz procs.

    I mean, good lord! If we compared those 7900GT numbers you linked to on GamePyre to the 8x.xx driver scores on SPEC's site, the 7900 GT is running very close to the HP Workstation (Opteron 256 3.0GHz) with 2X Quadro 4500's... 24.08 vs 20-25 (depending on which CPU you choose).

    This isn't working to compare numbers like these. I've come to the conclusion that SPECViewPerf 9 is seriously innacurate in measuring real performance difference.
     
    Last edited: Aug 16, 2006
  9. upf

    upf New Member

    Messages:
    7
    Likes Received:
    0
    GPU:
    6800GT > Quadro 4000
    SPECapc for MAYA 6 test results

    <TESTRUN>
    7.23,wireframe
    10.45,smoothShaded
    22.02,Texture Load Time
    14.97,textured
    15.2,texturedSelected
    24.24,texturedHilited
    scene,werewolf.ma
    4.12,Project File Load
    7.34,wireframe
    10,smoothShaded
    18.55,Texture Load Time
    15,textured
    15.44,texturedSelected
    24.55,texturedHilited
    scene,werewolf.ma
    4.2,Project File Load
    7.39,wireframe
    10,smoothShaded
    18.71,Texture Load Time
    15,textured
    15.58,texturedSelected
    24.67,texturedHilited
    scene,werewolf.ma
    4.28,Project File Load
    7.43,wireframe
    10,smoothShaded
    18.76,Texture Load Time
    14.98,textured
    15.59,texturedSelected
    24.71,texturedHilited
    scene,hand1.ma
    34.3,Project File Load
    2.36,wireframe
    3.97,smoothShaded
    0.05,Texture Load Time
    3.98,textured
    5.27,texturedSelected
    5.47,texturedHilited
    scene,hand1.ma
    33.22,Project File Load
    2.36,wireframe
    3.96,smoothShaded
    0.04,Texture Load Time
    3.97,textured
    5.24,texturedSelected
    5.46,texturedHilited
    scene,hand1.ma
    33.76,Project File Load
    2.36,wireframe
    3.99,smoothShaded
    0.06,Texture Load Time
    3.96,textured
    5.24,texturedSelected
    5.49,texturedHilited
    scene,Insect.ma
    3.25,Project File Load
    53.11,wireframe
    82.36,smoothShaded
    8.18,Texture Load Time
    183.02,textured
    198.76,texturedSelected
    216.96,texturedHilited
    scene,Insect.ma
    2.94,Project File Load
    53.28,wireframe
    82.58,smoothShaded
    7.5,Texture Load Time
    183.17,textured
    211.2,texturedSelected
    218.98,texturedHilited
    scene,Insect.ma
    2.97,Project File Load
    53.81,wireframe
    82.66,smoothShaded
    7.64,Texture Load Time
    183.24,textured
    203.22,texturedSelected
    220.52,texturedHilited
    scene,Squid.ma
    3.16,Project File Load
    7.78,wireframe
    15.46,smoothShaded
    35.3,Texture Load Time
    20.89,textured
    21.2,texturedSelected
    23.19,texturedHilited
    scene,Squid.ma
    2.96,Project File Load
    7.89,wireframe
    15.47,smoothShaded
    33.21,Texture Load Time
    21.4,textured
    21.94,texturedSelected
    23.19,texturedHilited
    scene,Squid.ma
    3,Project File Load
    8.02,wireframe
    15.47,smoothShaded
    33.41,Texture Load Time
    21.27,textured
    20.02,texturedSelected
    23.21,texturedHilited
    scene,handDef.ma
    33.97,Project File Load
    1333.59,smoothShaded
    </TESTRUN>
     
  10. augie

    augie Active Member

    Messages:
    58
    Likes Received:
    0
    GPU:
    Geforce 7900GT @ 630MHz
    Interesting numbers... Here are my numbers... I think I need to run the Quadro 3000 mode tests again... I just run it 1 time instead of the default 3 times. Just put a (1) instead of a (3) on the MEL script.

    Strangely, I have higher scores than you in most of the benchmarks using my 5900 Ultra in Geforce mode. CPU differences? Possibly. I cannot account for the HORRIBLE scores in Insect, Squid, and handDef on mine in Quadro 3000 mode.

    I'll run these a couple more times, and maybe try a 9x.xx driver in Geforce mode. (I'm stuck at 45.28 driver for Quadro mode on this card.)


    P4@3.3GHz - 5900 U (42.58 driver)
    P4@3.3GHz - SQ FX3000 (42.58 driver)
    Pentium D 820 2.8 Ghz - SQ FX4000 (91.36 drivers) Edit: updated...

    wireframe 6.63 6.83 7.23
    smoothShaded 10.18 10.31 10.45
    Texture Load Time 15.69 19.61 22.02
    textured 15.11 19.32 14.97
    texturedSelected 20.23 25.85 15.2
    texturedHilited 32.86 42.44 24.24
    werewolf.ma
    Project File Load 7.06 11.55 4.12

    wireframe 10.71 14.79 7.43
    smoothShaded 20.72 23.23 10
    Texture Load Time 30.16 33.92 18.76
    textured 33.37 35.22 14.98
    texturedSelected 42.25 41.51 15.59
    texturedHilited 55.73 58.51 24.71
    hand1.ma
    Project File Load 77.96 91.67 33.76

    wireframe 3.6 3.88 2.36
    smoothShaded 9.03 10.18 3.99
    Texture Load Time 0.15 0.14 0.06
    textured 9.05 10.16 3.96
    texturedSelected 11.92 13.2 5.24
    texturedHilited 13.36 14.83 5.49
    Insect.ma
    Project File Load 7.81 9.56 3.25

    wireframe 133.42 151.21 53.81
    smoothShaded 241.69 258.75 82.66
    Texture Load Time 12.29 11.73 7.64
    textured 552.49 232.76 183.24
    texturedSelected 567.41 2.86 203.22
    texturedHilited 633.78 2.89 220.52
    Squid.ma
    Project File Load 7.2 8.12 3.16

    wireframe 19.78 0.69 8.02
    smoothShaded 36.93 0.7 15.47
    Texture Load Time 69.3 0 33.41
    textured 52.41 0.71 21.27
    texturedSelected 51.95 0.69 20.02
    texturedHilited 68.59 0.7 23.21
    handDef.ma
    Project File Load 78.65 79.22 33.97

    smoothShaded 3220.5 0.09 1333.59
     
    Last edited: Aug 16, 2006

  11. upf

    upf New Member

    Messages:
    7
    Likes Received:
    0
    GPU:
    6800GT > Quadro 4000
    The numbers are in Quadro mode with 91.36 driver. All other versions of drivers(6... 7.... 8.... ) shows Weighted Geometric Mean = ~17.


    Card release date is october 2004. Is this some advantage?
     
    Last edited: Aug 16, 2006
  12. clokkevi

    clokkevi Master Guru

    Messages:
    368
    Likes Received:
    0
    GPU:
    GeForce GTX 680
    Very, very true.
    I seem to recall that Unwinder made a test with some older version of SPECviewperf.
    What he did was
    1. run the tests in normal mode
    2. tweak the test, in that he made a very few (like, 10) pixels be changed while running the tests.
    3. compare the results.

    The differences were - I don't remember, but I believe like 200% or more.

    His theory was that:

    The NVidia drivers have built in functions that recognize the pictures in the SPECviewperf tests.
    When they "see" such a picture, they go, like "AHA! Someone is running SPECviewperf! Then, we can just do this little trick...there, and.. there, and now we will make the test 200% faster!"
    And so, when he made just 10 pixels of the pictures different,
    the NVidia drivers did not recognize them, failed to spot that a SPECviewperf test was being performed.. and just went on in their usual way.

    The bad thing is that ATI most likely ALSO do this with their FireGL drivers.

    So as a conclusion, only self-made tests are 100% sure to be non-optimized.
    SPECopc for Maya /3ds Max etc.. these are as good as you can get without creating your own test, as they are using the actual application.


    Therfore, I find SPECviewperf useful for 1 thing only,
    and that is to check if a Quadro mod is working or not.

    I always check with the ugs-04 test in SPECviewperf 8.1,
    because that one is so extremely different.

    E.g. here are my results

    Code:
    ugs-04 Viewset - GeForce 6800 GT mode:
    -------------------------------------------------------------------------------
    Test Weight  Frames    DLB  Visual Double     Frame Buffer     Depth  Stencil  
      #    Per Sec   Sec   ID    Buffer Red Green Blue Alpha Buffer  Buffer  
    -------------------------------------------------------------------------------
      1   16.50   2.72     N/A   7      True    8    8    8    0     24      0   
      2   16.50   2.94     N/A   7      True    8    8    8    0     24      0   
      3   16.50   2.49     N/A   7      True    8    8    8    0     24      0   
      4   16.50   2.71     N/A   7      True    8    8    8    0     24      0   
      5    8.50   32.9     N/A   7      True    8    8    8    0     24      0   
      6    8.50   41.1     N/A   7      True    8    8    8    0     24      0   
      7    8.50   4.08     N/A   7      True    8    8    8    0     24      0   
      8    8.50   7.63     N/A   7      True    8    8    8    0     24      0   
    -------------------------------------------------------------------------------
    ugs-04 Weighted Geometric Mean =   4.774

    Code:
    ugs-04 Viewset - Quadro FX 4000 mode:
    -------------------------------------------------------------------------------
    Test Weight  Frames    DLB  Visual Double     Frame Buffer     Depth  Stencil  
      #    Per Sec   Sec   ID    Buffer Red Green Blue Alpha Buffer  Buffer  
    -------------------------------------------------------------------------------
      1   16.50   28.9     N/A   7      True    8    8    8    0     24      0   
      2   16.50   33.0     N/A   7      True    8    8    8    0     24      0   
      3   16.50   26.0     N/A   7      True    8    8    8    0     24      0   
      4   16.50   29.4     N/A   7      True    8    8    8    0     24      0   
      5    8.50   47.6     N/A   7      True    8    8    8    0     24      0   
      6    8.50   58.9     N/A   7      True    8    8    8    0     24      0   
      7    8.50   44.5     N/A   7      True    8    8    8    0     24      0   
      8    8.50   47.6     N/A   7      True    8    8    8    0     24      0   
    -------------------------------------------------------------------------------
    ugs-04 Weighted Geometric Mean =   34.92
    If the results are this different, I know the Quadro mod is working.


    Ahem, GeForce FX 5900 - Quadro FX 3000 will NOT show this difference in this test.
    This is because those are NOT 100% equal.
    On those I instead use the first test, 3dsmax-03.
     
    Last edited: Aug 17, 2006
  13. CodemasterPrime

    CodemasterPrime New Member

    Messages:
    4
    Likes Received:
    0
    GPU:
    NVidia XFX 6800GT 256M
    I have an odd mix actually ...

    this is on an AMD 64 3700 'San Diego' core (2.2Ghz core speed), ASUS SLI motherboard (but 1 video card), 4G of RAM but only 3G useable (32 bit mode), Maya 6.5.

    My C drive was near full so fragmentation may have affected it.

    The two runs here were both without changing the video driver, which is 9131. (whatever I had most recently upgraded to).

    Numbers on the right is the Softquad setting, on the left is the 'stock' 6800GT and driver.

    I did get an error on the first run, I forget exactly but I had to click an 'ok' button so this may have affected the first test.

    The 'insect' tests are odd, 3 runs and on the stock 6800GT there was a huge variability. Given my hard drive was nearly full I wonder if this was an abberation.

    Code:
    				6800GT	
    				Normal	SoftQuad
    	wireframe			1.21	8.56
    	smoothShaded			0.02	0.09
    	Texture	Load	Time	0	0
    	textured			0.01	0.08
    	texturedSelected			0.02	0.08
    	texturedHilited			0.01	0.09
    werewolf.ma					
    	Project	File	Load	1.83	1.84
    	wireframe			1.21	8.56
    	smoothShaded			0.02	0.08
    	Texture	Load	Time	0	0.01
    	textured			0.01	0.08
    	texturedSelected			0.02	0.08
    	texturedHilited			0.02	0.08
    werewolf.ma					
    	Project	File	Load	0.7	0.86
    	wireframe			1.19	8.54
    	smoothShaded			0.01	0.09
    	Texture	Load	Time	0.02	0
    	textured			0.01	0.1
    	texturedSelected			0	0.08
    	texturedHilited			0	0.07
    werewolf.ma					
    	Project	File	Load	0.72	0.61
    	wireframe			1.2	8.54
    	smoothShaded			0.01	0.09
    	Texture	Load	Time	0	0
    	textured			0.02	0.08
    	texturedSelected			0.01	0.07
    	texturedHilited			0.02	0.08
    hand1.ma					
    	Project	File	Load	36.71	37.92
    	wireframe			3.54	0.01
    	smoothShaded			10.66	0
    	Texture	Load	Time	0.14	0
    	textured			10.67	0
    	texturedSelected			13.2	0
    	texturedHilited			15	0
    hand1.ma					
    	Project	File	Load	36.33	36.73
    	wireframe			3.55	0.02
    	smoothShaded			10.69	0
    	Texture	Load	Time	0.16	0
    	textured			10.67	0
    	texturedSelected			13.11	0.01
    	texturedHilited			14.89	0
    hand1.ma					
    	Project	File	Load	36.89	37.42
    	wireframe			3.54	0
    	smoothShaded			10.64	0
    	Texture	Load	Time	0.14	0
    	textured			10.65	0.01
    	texturedSelected			13.13	0
    	texturedHilited			14.77	0
    Insect.ma					
    	Project	File	Load	2.24	2.37
    	wireframe			21.69	0.72
    	smoothShaded			38.5	0.7
    	Texture	Load	Time	1.56	0
    	textured			51.97	0.72
    	texturedSelected			59.52	0.69
    	texturedHilited			61.85	0.7
    Insect.ma					
    	Project	File	Load	2.06	2.06
    	wireframe			21.69	0.72
    	smoothShaded			38.8	0.71
    	Texture	Load	Time	1.19	0
    	textured			53.2	0.72
    	texturedSelected			22.22	0.69
    	texturedHilited			0.72	0.72
    Insect.ma					
    	Project	File	Load	2.08	2.07
    	wireframe			0.7	0.72
    	smoothShaded			0.7	0.72
    	Texture	Load	Time	0	0
    	textured			0.72	0.74
    	texturedSelected			0.69	0.7
    	texturedHilited			0.7	0.72
    Squid.ma					
    	Project	File	Load	2.07	2.44
    	wireframe			0.1	0.09
    	smoothShaded			0.09	0.09
    	Texture	Load	Time	0	0
    	textured			0.11	0.11
    	texturedSelected			0.09	0.1
    	texturedHilited			0.1	0.1
    Squid.ma					
    	Project	File	Load	2.03	2.05
    	wireframe			0.09	0.09
    	smoothShaded			0.1	0.09
    	Texture	Load	Time	0	0
    	textured			0.09	0.11
    	texturedSelected			0.09	0.1
    	texturedHilited			0.09	0.09
    Squid.ma					
    	Project	File	Load	2.06	2.05
    	wireframe			0.09	0.09
    	smoothShaded			0.1	0.09
    	Texture	Load	Time	0	0
    	textured			0.1	0.11
    	texturedSelected			0.09	0.1
    	texturedHilited			0.11	0.09
    handDef.ma					
    	Project	File	Load	13.57	13.37
    	smoothShaded			0.01	0.02
    					
    So .. what do you think?

    I'm going to try World of Warcraft :) and see what FPS I get, then try downgrading the drivers to the last supported SoftQuad drivers.

    == John ==
     
  14. augie

    augie Active Member

    Messages:
    58
    Likes Received:
    0
    GPU:
    Geforce 7900GT @ 630MHz
    Your results are very odd indeed... Hand and insect, the Geforce mode blows away the Quadro mode by a ridiculous margin. Warewolf shows the Quadro clearly in the lead, esp in wireframe.

    I have to assume something went wrong on your Quadro mode tests. How could there be so many 0's ? I don't think there should be any 0's in there...

    I've already got my hands on a 6800 Ultra, but alas no PCI-E mobo yet (building a new Core 2 Duo system). I'm trying to buy a used 7900 GT as well, so this question (my thread name) will be answered once and for all within a couple of weeks. :O

    The answer didn't come cheap. :p

    Augie
     
  15. Stefan_VTi

    Stefan_VTi Member

    Messages:
    47
    Likes Received:
    0
    GPU:
    PNY Quadro FX 3500
    Back again with some more info! After the weird results I decided to write a small benchmark macro for SolidWorks, so I could test the actual application performance. I tested 2 drivers, 84.26 and 91.36. I tested both with high performance settings and with 8xs AA and all other eye candy on.

    My benchmark script tests rotating & zoom performance over 1000 iterations, results are in seconds required (lower is better).

    Driver 84.26 - Performance settings
    1000 Rotate actions: 83.0
    1000 Zoom actions: 89.1

    Driver 84.26 - Quality settings
    1000 Rotate actions: 84.6
    1000 Zoom actions: 137.2

    Driver 91.36 - Performance settings
    1000 Rotate actions: 82.4
    1000 Zoom actions: 94.7

    Driver 91.36 - Quality settings
    1000 Rotate actions: 84.2
    1000 Zoom actions: 139.5

    Not quite what you would expect when you look at the specviewperf results of both drivers...
     

  16. CodemasterPrime

    CodemasterPrime New Member

    Messages:
    4
    Likes Received:
    0
    GPU:
    NVidia XFX 6800GT 256M
    Ok, I backdated my driver to the reference / certified driver, softmodded, and tried the test again. I still get that error by the way, a ; missing, on the first run.

    I also figured out how to format the results better with the .xls file. Here's the results; Forceware 81.67 (last certified Maya drivers),
    6800GT softmodded to an FX4000:


    Code:
    	Run 1	Run 2	Run 3	Avg	Ratio		
    werewolf.ma							
    Project	0.73	0.76	0.69	0.73	7.62	GFX	63.71
    wireframe	8.14	8.13	8.14	8.14	0.61	I/O	100.25
    smoothShaded	0.06	0.08	0.08	0.07	115.91		
    Texture	0.02	0.01	0.01	0.01	1319.00		
    textured	0.07	0.07	0.07	0.07	192.86		
    texturedSelected	0.06	0.06	0.06	0.06	259.67		
    texturedHilited	0.06	0.06	0.06	0.06	298.89		
    							
    	Run 1	Run 2	Run 3	Avg			
    hand1.ma							
    Project	35.42	37	36.9	36.44	1.70	GFX	0.34
    wireframe	29.22	29.48	0.02	19.57	0.23	I/O	0.73
    smoothShaded	56.43	55.99	0	37.47	0.33		
    Texture	0.82	0.84	0	0.55	0.31		
    textured	56.3	56.03	0.01	37.45	0.33		
    texturedSelected	83.17	75.73	0	52.97	0.31		
    texturedHilited	99.08	0.02	0.02	33.04	0.54		
    							
    	Run 1	Run 2	Run 3	Avg			
    Insect.ma							
    Project	2.11	2.06	2.08	2.08	1.97	GFX	165.60
    wireframe	0.71	0.72	0.72	0.72	68.56	I/O	Err:503
    smoothShaded	0.71	0.7	0.71	0.71	119.75		
    Texture	0	0	0	0.00	Err:503		
    textured	0.71	0.72	0.72	0.72	235.00		
    texturedSelected	0.7	0.7	0.71	0.70	259.29		
    texturedHilited	0.72	0.74	0.72	0.73	248.93		
    							
    	Run 1	Run 2	Run 3	Avg			
    Squid.ma							
    Project	2.15	2.07	2.05	2.09	1.92	GFX	190.57
    wireframe	0.09	0.1	0.09	0.09	110.96	I/O	Err:503
    smoothShaded	0.11	0.09	0.11	0.10	160.58		
    Texture	0	0	0	0.00	Err:503		
    textured	0.09	0.09	0.09	0.09	246.48		
    texturedSelected	0.11	0.11	0.1	0.11	226.69		
    texturedHilited	0.09	0.1	0.11	0.10	252.43		
    							
    	Run 1	Run 2	Run 3	Avg			
    handDef.ma							
    Project	13.54	13.54	13.54	13.54	1.75	CPU	84993.00
    smoothShaded	0.01	0.01	0.01	0.01	84993.00	I/O	1.75
    
    The 0's are weird ... I wonder if there's a bug that prevents the graphics display?

    Here's my initial, unsoftmodded run:
    Code:
    	Run 1	Run 2	Run 3	Avg	Ratio		
    werewolf.ma							
    Project	1.83	0.7	0.72	1.08	5.11	GFX	353.87
    wireframe	1.21	1.19	1.2	1.20	4.10	I/O	116.11
    smoothShaded	0.02	0.01	0.01	0.01	637.50		
    Texture	0	0.02	0	0.01	2638.00		
    textured	0.01	0.01	0.02	0.01	1012.50		
    texturedSelected	0.02	0	0.01	0.01	1558.00		
    texturedHilited	0.02	0	0.02	0.01	1345.00		
    							
    	Run 1	Run 2	Run 3	Avg			
    hand1.ma							
    Project	36.71	36.33	36.89	36.64	1.69	GFX	1.21
    wireframe	3.54	3.55	3.54	3.54	1.25	I/O	1.41
    smoothShaded	10.66	10.69	10.64	10.66	1.17		
    Texture	0.14	0.16	0.14	0.15	1.18		
    textured	10.67	10.67	10.65	10.66	1.17		
    texturedSelected	13.2	13.11	13.13	13.15	1.27		
    texturedHilited	15	14.89	14.77	14.89	1.20		
    							
    	Run 1	Run 2	Run 3	Avg			
    Insect.ma							
    Project	2.24	2.06	2.08	2.13	1.93	GFX	4.94
    wireframe	21.69	21.69	0.7	14.69	3.34	I/O	3.14
    smoothShaded	38.5	38.8	0.7	26.00	3.25		
    Texture	1.56	1.19	0	0.92	5.11		
    textured	51.97	53.2	0.72	35.30	4.77		
    texturedSelected	59.52	22.22	0.69	27.48	6.64		
    texturedHilited	61.85	0.72	0.7	21.09	8.58		
    							
    	Run 1	Run 2	Run 3	Avg			
    Squid.ma							
    Project	2.07	2.03	2.06	2.05	1.96	GFX	195.63
    wireframe	0.1	0.09	0.09	0.09	110.96	I/O	Err:503
    smoothShaded	0.09	0.1	0.1	0.10	171.66		
    Texture	0	0	0	0.00	Err:503		
    textured	0.11	0.09	0.1	0.10	221.83		
    texturedSelected	0.09	0.09	0.09	0.09	268.67		
    texturedHilited	0.1	0.09	0.11	0.10	252.43		
    							
    	Run 1	Run 2	Run 3	Avg			
    handDef.ma							
    Project	13.57	13.57	13.57	13.57	1.74	CPU	84993.00
    smoothShaded	0.01	0.01	0.01	0.01	84993.00	I/O	1.74
    

    == John ==
     
  17. CodemasterPrime

    CodemasterPrime New Member

    Messages:
    4
    Likes Received:
    0
    GPU:
    NVidia XFX 6800GT 256M
    Bump, and a quote from higher up:

    That's probably true, but I get the rather vauge impression that Maya isn't quite as bad about wanting the Quadro features to increase speed.

    So I'm mainly interested in actual Maya benchmarks.

    Has anyone tried a 7900 of one kind or another with Maya?

    Your results above make me think I should just use my 6800 instead of upgrading, but I need a 512Meg card ... and I'd like the performance upgrade for other stuff.

    == John ==
     
  18. augie

    augie Active Member

    Messages:
    58
    Likes Received:
    0
    GPU:
    Geforce 7900GT @ 630MHz
    I did end up buying a 6800 Ultra and a 7900 GT SC (overclocked) to test, but I have been so busy with other things and trying to get this new Core2 system optimized I've only done some very preliminary tests. I'll try to post some data soon, but for now, basically I can't see any difference in actual Maya use. I'm kind of unsure what individual processes I should test like high-poly modeling, real time lighting, textured polys, high-quality rendering, hardware render, sculpt surface tools, etc... If you have any requests, please let me know.

    I have been doing a lot of game level design in Maya lately, which is mainly high-poly textured scenes -- benchmarks on my game level shows NO performance difference whatsoever. I'm assuming my limit is more CPU based than video card based in this case.

    In terms of benchmarks, there is definitely a large and noticable difference using Spec ViewPerf 9 between the 6800U Quadro and the 7900 GT SC. The Quadro wins by a large margin on some of the benchmarks. But again, I haven't seen this in my Maya work.

    I'll run the SpecViewPerf 8 and 9 benches, then the SpecCapc for Maya 6.0. Can anybody think of any other ways to bench Maya so we can figure this out?

    Augie
     
  19. deathvirus

    deathvirus Ancient Guru

    Messages:
    4,507
    Likes Received:
    13
    GPU:
    RTX 2070 MaxQ 8GB
    I might agree a bit on this .... did face a similar thing with my older card ... Max does though ...
     
  20. Samiam

    Samiam Member Guru

    Messages:
    144
    Likes Received:
    0
    GPU:
    4G DDR2
    7900 Maya testing suggestions

    Hi Augie,

    Here's how i benchmark Quadro/game card capabilities:

    I playback looped camera spins of my models/scenes in Maya, timing them with a stopwatch. That's accurate enough to detect signifigant changes in performance between video cards and reflects real world performance. To save time benchmarking I enable several rendering features in each test pass to look for changes big enough between cards to merit more investigation. I find Maya is heavilly CPU/system dependant.

    I'll do a wireframe and shaded test of course. Another test will include advanced features like 2-sided lighting, high quality lighting mode, normal mapping, and shadows enabled.

    Artisan tools are slower and have visual artifacts on older game cards without overlays. Perhaps the 7900 has enough brute force ovrlays aren't necessary? An easy way to test that is gradually increasing the complexity of a scene on a quadro setup, saving incrementally, while painting vertex values, textures and such. Can be as simple as a distorted tesselated sphere, maybe with some bones or heavy texturing or something. As your scene begins to slow down, revert to last save with smooth performance and test that scene on the game setup. That'll quickly tell if the game card chugs or not and if the card meets your production needs.

    You could also test general image quality. Lack of overlays is supposed to create menu artifacts.

    Another good test would be hardware assisted rendering.

    I'd be curious about your results. I'm considering upgrading but with no intention of buying Quadro. The era of acceptability of SGI-like prices for marginally different "professional" 3D hardware is long past imho.

    Consumers should say "enough" and Maya should better support powerful mainstream video cards. Maya reccomends a video card which may be more expensive than either the software or the rest of the CPU, and which yeilds only marginal improvements! How crazy is that? :bonk:

    Anyways, good luck with tests! Let us know. :)
     
    Last edited: Oct 17, 2006

Share This Page