Search Unity

  1. Megacity Metro Demo now available. Download now.
    Dismiss Notice
  2. Unity support for visionOS is now available. Learn more in our blog post.
    Dismiss Notice

Unity Benchmark v0.1

Discussion in 'General Discussion' started by Arowx, Sep 20, 2012.

  1. Toasttify

    Toasttify

    Joined:
    Mar 27, 2012
    Posts:
    143
    On the 4.1 version...

    Fast to : 16,305
    Slow to: 19,545
    Amd A10 5800k
    Gtx 480
    8 gigs ram

    And just tested this on my 2 core 1ghz netbook...
    3454 cubes, not bad for a netbook haha.
     
    Last edited: Mar 16, 2013
  2. Arowx

    Arowx

    Joined:
    Nov 12, 2009
    Posts:
    8,194
    Last edited: Mar 17, 2013
  3. Birdlay

    Birdlay

    Joined:
    Jul 28, 2012
    Posts:
    39
    GPU test v0.2: 15642 cubes :) Surprisingly good for an Asus 9600gt 512MB from what... 2006? :D
     
  4. SpookyCat

    SpookyCat

    Joined:
    Jan 25, 2010
    Posts:
    3,748
    Getting a score of 39750 cubes on Asus 6870 card and i7 3.6Ghz Quad Core.
     
  5. MarigoldFleur

    MarigoldFleur

    Joined:
    May 12, 2012
    Posts:
    1,353
    18,495 on my 2009 13 inch base level MacBook Pro.
     
  6. Arowx

    Arowx

    Joined:
    Nov 12, 2009
    Posts:
    8,194
    Wow it looks like the Intel CPU really shines on this benchmark maybe Unity could spend some time with AMD to try and get a bit more out of their processors [jealous AMD owner]?

    Just checked and it looks like AMD are down to about a quarter of the market share (source Unity and Steam stats).

    Thinking of adding a simple set of benchmarks, and UI to this so you can select to test with various sets of colliders/shaders/physics etc would that work, and then allow you to run the whole thing in benchmark mode and provide a total score?

    So is there anything you would like me to add to this benchmark?
     
  7. CharlieSamways

    CharlieSamways

    Joined:
    Feb 1, 2011
    Posts:
    3,424
    installed another 4gb of ram, thought i had 20, turned out I only had my 16 in! got to 32k
     
  8. ColossalDuck

    ColossalDuck

    Joined:
    Jun 6, 2009
    Posts:
    3,246
    28500 and it was still climbing.
     
  9. Arowx

    Arowx

    Joined:
    Nov 12, 2009
    Posts:
    8,194
    OK I've had a little play with a new Menu for Cube Mark, might have gotten a bit carried away, but what do you think!?

    $screenShotMenu.png

    Note may need shader model 3+ graphics cards, and stress the GPU more than the benchmark! ;0)
     
  10. Arowx

    Arowx

    Joined:
    Nov 12, 2009
    Posts:
    8,194
    $screenShot_v04.png

    OK CubeMark v0.4 (click image) with flashy intro screen, which may turn into interface.

    New features

    • Quad Mark
    • Cube Mark
    • Cube Mark + Colliders
    • Run All in Batch Mode
    • System Stats
    • Record of Last CubeMarks

    TIP Open the Webplayer in fullscreen mode (right click for menu) for higher marks.
     
    Last edited: Mar 21, 2013
  11. Arowx

    Arowx

    Joined:
    Nov 12, 2009
    Posts:
    8,194
    Unity 4.1 Score


    Benchmarks
    2D Mark: 7011
    3D No Colliders: 19300
    3D Colliders: 6705


    Unity 4.1.1


    Benchmarks [Browser]
    2D Mark: 10075
    3D No Colliders: 20105
    3D Colliders: 6640


    Benchmarks [Fullscreen #1]
    2D Mark: 10310
    3D No Colliders: 20045
    3D Colliders: 6705


    Benchmarks [Fullscreen #2]
    2D Mark: 10290
    3D No Colliders: 20420
    3D Colliders: 6640

    4.1.1 seems to be a bit faster but note there is over 10% variation on some benchmark runs.

    And I need to look into why the 2D version which just replaces the Cube with a Quad is way slower than the 3D?

    TIP: You can copy the stats text
     
  12. ZJP

    ZJP

    Joined:
    Jan 22, 2010
    Posts:
    2,649
    Benchmarks[Browser]

    2D Mark: 8220
    3D No Colliders: 17435
    3D Colliders: 5480

    HD 7850, C2Q Q8300 @ 3ghz
     
  13. Dinoboy626

    Dinoboy626

    Joined:
    Dec 9, 2012
    Posts:
    1
    7720 cubes
    Target: 30 fps
    GPU: Mobile Intel 4 Series Express Chipset Family
    CPU: Intel Core 2 Duo @ 2.00 GHz

    No settings changed with everything i normally have open opened.

    Note: I did this on my nearly 3 year old laptop.



    Note: There was no overclocking involved.
     
    Last edited: Mar 22, 2013
  14. Arowx

    Arowx

    Joined:
    Nov 12, 2009
    Posts:
    8,194
    Quick Update fixed the 2D Quad problem

    And added total CubeMark to results.

    Also worked on making the benchmark run a bit faster.

    CubeMark v0.4

    Also set up Facebook page if you want to post your top marks and share with friends

    CubeMark on Facebook
     
    Last edited: Mar 22, 2013
  15. Arowx

    Arowx

    Joined:
    Nov 12, 2009
    Posts:
    8,194
    Wow just built a standalone version of Cube Mark and got this score at full monitor resolution

    That's 149.7% faster!

    But would anyone be interested in buying a PC version of a free web app?
     
  16. Arowx

    Arowx

    Joined:
    Nov 12, 2009
    Posts:
    8,194
    Adding an Oculus Rift mode to the benchmark, still a few bugs to work out as it will also need a 60fps target.

    I don't have the distortion pattern required for the Rift at the moment but it was very easy to add stereoscopic cameras in Unity.

    $screenShot_v04riftMode.png

    Once it's done I will update the online webplayer version.
     
  17. Arowx

    Arowx

    Joined:
    Nov 12, 2009
    Posts:
    8,194
    OK New version with:


    • Oculus Rift Mode*
    • 60 FPS Target Option
    • DX11 Tesselated Cubes

    *Well Sterioscopic and at Rift resolution and ideal VR 60fps target.

    For me at least the 60FPS target gets abysmal results unless I'm in full screen mode (right mouse menu).
     
  18. Ghoxt

    Ghoxt

    Joined:
    Mar 25, 2008
    Posts:
    104
    I installed a new Geforce Titan video card yesterday. The thing zipped right along. My machine is not a good testbed for development benchmarking (today) for all the obvious reasons. (see below)

    [edit] What i mean is that it's not a good gauge of most users hardware.

    2D Mark: 37244
    3D No Colliders: 52252
    3D Colliders: 9624
    Cube Mark: 99120

    $Cubemark Ghoxt High Quality mode 1.jpg
     
    Last edited: Mar 24, 2013
  19. Arowx

    Arowx

    Joined:
    Nov 12, 2009
    Posts:
    8,194
    Do you mean that the test is browser based so limited in it's maximum potential e.g. Browser Plugin / Resolutions?
    That the benchmark only uses cubes so low polygon throughput?
    That the benchmark does not stress high end GPU's enough?
    That the benchmark does not stress high end CPU's enough?
    That the benchmark does not represent a modern gaming experience, well apart from maybe an exploded view of minecraft?
     
  20. Ghoxt

    Ghoxt

    Joined:
    Mar 25, 2008
    Posts:
    104
    I actually didn't think along the lines of the Benchmark as limited at all, which I love btw. To be specific, I was meaning that 99% of the people won't have machines with capable video cards of 6Meg VRAM at 120fps and SSD's in Quad Raid-0, and that I need to make sure when I develop a game etc, that I make something that tests well on normal machines, and not a specialty build gaming rig.

    I'm shooting for Win / Mac / iPad / iPhone eventually as targets.

    PS. I edited my original post as I can see how it could have been misinterpreted as I was not specific. :D
     
  21. inafield

    inafield

    Joined:
    Feb 26, 2013
    Posts:
    281
    Browser Window @ 30fps
    2D Mark: 22318
    3D No Colliders: 28622
    3D Colliders: 6845
    Cube Mark: 57785

    Dell Latitude E6520
    Windows 7 SP1 64bit
    Intel i7-2640M @ 2.8GHz (it's actually dual core with hyperthreading, not quad core)
    8GB Ram
    Intel HD 3000 with 128Mb

    Asus G75VX
    2D: 25494
    3D No COlliders: 34376
    3D Colliders: 6980
    Cube Mark: 66850

    Win 8
    i7-3630QM
    Mem: 12 GB
    GeForce 670MX 3GB
     
    Last edited: Jul 22, 2013
  22. Arowx

    Arowx

    Joined:
    Nov 12, 2009
    Posts:
    8,194
    OK New version of Unity so updated the benchmark, standalone version, Unity Cube Mark 4.2

    The web player version is not working at the moment, probably due to the Oculus Rift SDK integration.
     
  23. inafield

    inafield

    Joined:
    Feb 26, 2013
    Posts:
    281
    Either 4.2 gave some massive speed boost, or there is a massive performance difference between browser and stand-alone.

    Asus G75VX, stand alone, 1280x720, windowed, good.
    2D: 36400
    3D No Colliders: 47465
    3D Colliders: 9909
    Cube Mark: 93774
     
  24. Arowx

    Arowx

    Joined:
    Nov 12, 2009
    Posts:
    8,194
    If you would like to compare it to the Unity 4.1 build give it a whirl.
     
    Last edited: Jul 22, 2013
  25. Arowx

    Arowx

    Joined:
    Nov 12, 2009
    Posts:
    8,194
  26. carl010010

    carl010010

    Joined:
    Jul 14, 2010
    Posts:
    139
    web browser
    OS : Mac 10.8.3(note I am using a hackintosh)
    CPU: Intel Core i7-3770 @ 4.20GHz MEM: 16.0 Gb (quad core)
    GPU NVIDIA GeForce GT 640 Open GL Engine MEM: 2.0Gb

    Benchmarks
    2D: 38012
    3D: 49559
    3D Colliders: 6855
    Cube Mark 90426

    Is this good?
     
    Last edited: Jul 26, 2013
  27. Arowx

    Arowx

    Joined:
    Nov 12, 2009
    Posts:
    8,194
    That is nearly double my own score of about 57000, that's in full screen mode as well! ;0)
     
  28. Arowx

    Arowx

    Joined:
    Nov 12, 2009
    Posts:
    8,194
  29. RandAlThor

    RandAlThor

    Joined:
    Dec 2, 2007
    Posts:
    1,293
    Tested with the 4.2

    Build my own machiene (about a year old):

    2D: 37325
    3D: 45123
    3D Colliders: 10548
    Cube Mark 92996

    Tested with 4.3

    2D: 30248
    3D: 42916
    3D Colliders: 10656
    Cube Mark 83820
     
    Last edited: Nov 12, 2013
  30. Arowx

    Arowx

    Joined:
    Nov 12, 2009
    Posts:
    8,194
  31. mgear

    mgear

    Joined:
    Aug 3, 2010
    Posts:
    9,350
    CubeMark v0.4 webplayer always crashes during the first test..

    exe 4.2 (v0.5), 1280x800, good, windowed
    2D Mark: 34604
    3D No Colliders: 42122
    3D Colliders: 8648
    Cube Mark: 85374

    exe 4.3 (v0.5), 1280x800, good, windowed
    2D Mark: 27437
    3D No Colliders: 37502
    3D Colliders: 8625
    Cube Mark: 73564


    Operating System: Windows 7 Ultimate 64-bit
    Processor: Intel(R) Xeon(R) CPU E5-2650 v2 @ 2.60GHz (16 CPUs), ~2.6GHz
    Memory: 65536MB RAM
    Card name: NVIDIA Quadro K4000


    *need option to copy paste those cubemark values, instead of typing, maybe javascript alertbox?..Oo?
     
  32. TylerPerry

    TylerPerry

    Joined:
    May 29, 2011
    Posts:
    5,577
    wow, dat ram!
     
  33. Arowx

    Arowx

    Joined:
    Nov 12, 2009
    Posts:
    8,194
    Wow quite a system, good idea need a copy to clipboard button.
     
  34. Sanguine Jackal

    Sanguine Jackal

    Joined:
    May 7, 2013
    Posts:
    62
    Holy mother of Pete, I want some of that RAM. Jeebus!

    Sorry I have nothing constructive to add, just dayumn.
     
  35. Arowx

    Arowx

    Joined:
    Nov 12, 2009
    Posts:
    8,194
  36. Arowx

    Arowx

    Joined:
    Nov 12, 2009
    Posts:
    8,194
  37. lorenalexm

    lorenalexm

    Joined:
    Dec 14, 2012
    Posts:
    307
    2D Mark: 23078
    3D No Colliders: 33361
    3D Colliders: 8228
    Cube Mark: 64667

    OS: Mac OS X 10.9.2
    CPU: Intel(R) Core(TM) i5-3210M CPU @ 2.50GHz
    MEM: 4.0 Gb
    GPU: Intel HD Graphics 4000 OpenGL Engine MEM: 1.0 Gb

    Not to shabby for my little MacBook Pro; even more-so when compared to the results given by mgear's beast of a machine.
     
  38. lmbarns

    lmbarns

    Joined:
    Jul 14, 2011
    Posts:
    1,628
    Lame....

    Unity 4.2:
    $unity42.jpg

    Unity 4.3:
    $unity43.jpg

    Big Difference.....


    Any chance at a mac build? I'd like to test my mac mini to see how bad it is on S***tier hardware....


    Here's in the web player link posted above:
    $webplayer.jpg

    And the system specs:
    $systemspecs.jpg

    Webplayer is so much faster wtf


    edit2:: Webplayer on my $1500 mac mini I got to use with wiibudy-
    2D MARK: 27167
    3D NO COLLIDER: 39366
    3D COLLIDER: 9015
    CUBE MARK: 75548

    SYSTEM:
    OS: MAC OS X 10.8.5
    CPU: INTEL(R) CORE(TM) I7-3615QM CPU @ 2.3GHZ MEM: 4GB
    GPU: INTEL HD GRAPHICS 4000 OPENGL ENGINE MEM: 1.1GB
     
    Last edited: Mar 3, 2014
  39. landon912

    landon912

    Joined:
    Nov 8, 2011
    Posts:
    1,579
    Thats a sexy computer. I prefer a GTX GPU for gaming though ;).

    CubeMark: 57,979
     
    Last edited: Mar 3, 2014
  40. Arowx

    Arowx

    Joined:
    Nov 12, 2009
    Posts:
    8,194
    Odd that tried it on a new PC build and I'm also getting way faster results on webplayer than PC. Thought it might be the quality settings and resolution so changed that. But although it was a bit faster still no-way as fast at the webplayer.

    4.3.4 build for PC - https://dl.dropboxusercontent.com/u/19148487/CubeMark/Unity4.3/CubeMark434PC.zip

    The only other difference is the webplayer version has the Oculus Rift features and dll's removed?

    Sorry I'm not setup to build to mac at the moment, have you tried the web player version on your mac?

    Forgot you can toggle the target frame rate to 30fp or 60fps (right hand spinning cube) this could be the problem.
     
    Last edited: Mar 3, 2014
  41. jonkuze

    jonkuze

    Joined:
    Aug 19, 2012
    Posts:
    1,709
    $1.png
    $2.png

    The Only Issue I have with my system and this Benchmark is that i'm not 100% sure if it's accurate, as I have a
    HP Pavilion g6t-2200 CTO Laptop which has Dual Graphics Cards the Intel HD and an Radeon HD 7670M it's suppose to auto switch betwen 3D Applications and 2D Applications. But I'm not so sure it works correctly...

    $3.png

    There is a way to Assign what Applications Run on which Graphic Card; I have Assigned Cube Mark, and Unity to Run on my Radeon HD 7670M but still it shows that i'm using Intel HD which makes no sense! Because lets say if I go into my Device Manager and Disable the Intel HD Display Adapter, to ensure i'm only running on Radeon, I can't seem to Run any 3D Applications at all... I get Errors from my Graphic Driver saying AMD Radeon is not Enabled when in Fact it is... So weird Crap going on with this Laptop.. ugh...
     
    Last edited: Mar 3, 2014
  42. lmbarns

    lmbarns

    Joined:
    Jul 14, 2011
    Posts:
    1,628
    Yeah I tested the first browser link on a mac mini and it smoked my dev machine running the standalone in both versions. Dev machine in browser was much faster than standalone, but yeah maybe it's more efficient compression or something for web. But it's dramatic between the two standalone versions on the same machine...screenshots above.
    4.2 cubemark was 43700 and 4.3 on same machine, standalone, was 35200....
     
  43. Arowx

    Arowx

    Joined:
    Nov 12, 2009
    Posts:
    8,194
    OK I've tried it again and managed to get a slightly faster standalone score when I set the screen size to 800 x 600 and quality to good (matching the webplayer version), then fps to 30 (slow spinning cube on right).

    There was an initial performance drop with 4.3 but I thought UT addressed that with a fix, was that using the 4.3.4 version?

    OK checked 4.2 and got 66k but 4.3.4 got 62.2k or a drop in performance of 0.5%.
     
    Last edited: Mar 4, 2014
  44. lmbarns

    lmbarns

    Joined:
    Jul 14, 2011
    Posts:
    1,628
    I'm on a diff computer now, will try again tomorrow but what's the big difference between your 4.3 and 4.3.4 version? Between the original 4.3 and 4.2 there was more than 12% drop for me....viewed at the same settings on the same computer.

    Then the webplayer was twice as fast as either, probably from the platform, but really? Twice as fast in the browser vs standalone on same comp......
     
  45. deram_scholzara

    deram_scholzara

    Joined:
    Aug 26, 2005
    Posts:
    1,043
    My new hackintosh (running in windows)
    $Hackintosh(Banshee).png
     
  46. Arowx

    Arowx

    Joined:
    Nov 12, 2009
    Posts:
    8,194

    There was a performance problem with the initial release of 4.3 that was resolved in an update, so in theory the 4.3.4 version should run a lot closer to the 4.2 version, for me there was only a 0.5% drop in performance (note I only did one run, ideally I should set it up to do multiple runs from a cold startup without any other apps running and provide an average).
     
  47. lmbarns

    lmbarns

    Joined:
    Jul 14, 2011
    Posts:
    1,628
    Ok so the new 4.3.4 is a lot faster than the original.

    $unity42.jpg vs $unity434.jpg

    Still a little slower.
     
  48. Arowx

    Arowx

    Joined:
    Nov 12, 2009
    Posts:
    8,194
    So about 4% slower, which is better than my rigs 5% slower score.
     
  49. Arowx

    Arowx

    Joined:
    Nov 12, 2009
    Posts:
    8,194
  50. PJisAnarchist

    PJisAnarchist

    Joined:
    Jan 9, 2012
    Posts:
    9
    The performance gap between 4.2 and 4.3 is a shame! Could it be something in CubeMark that changed or did you just built the exact same project?

    exe 4.2 (v0.5), 1280x800, good, windowed
    2D Mark: 42595
    3D No Colliders: 47834
    3D Colliders: 11664
    Cube Mark: 102093

    exe 4.3 (v0.5), 1280x800, good, windowed
    2D Mark: 34948
    3D No Colliders: 42210
    3D Colliders: 11444
    Cube Mark: 88602


    Operating System: Windows 7 Ultimate 64-bit
    Processor: Intel(R) Core i7-4930K CPU@ 3.4GHz (6 CPUs), 3.3GHz
    Memory:16GB RAM
    Card name: NVIDIA Titan Black Edition