SPECapc – Maya® 2012 GPU Scores

The SPECapc for Maya® 2012 is a benchmark developed by The Standard Performance Evaluation Corporation (SPEC). It is the updated version of the synonymous 2009 benchmark, utilizing the same models in most of the tests.

Instead of reporting the average frames per second the GPU is capable of rendering for each test, the SPECapc for Maya® 2012 benchmark computes composite scores for graphics and CPU performance, which are then normalized against a reference system that would score 1.0. The reference system is a Dell Precision 690 with Xeon 5130 2GHz, NVIDIA Quadro FX 570 GPU  and 16GB RAM in 4 x 4GB ECC DDR2 667MHz Dimms.

For this page, we won’t include CPU performance, as we are interested in the GPU performance evaluation. The test results for the cards tested so far are as follows:

pcfoo_Specapc_Maya2012-Q1_2014_600

Test Configuration

All cards above have been tested using our i7-3960K overclocked @ 4.6GHz (46×100), on an Asus P9X79 Pro and 16GB (4x4GB Samsung 30nm) Quad Channel  2333MHz DDR3.
Using overclocked, high-end CPUs and fast RAM, we are trying to minimize the CPU bottlenecks and keep the results relevant for a couple of years as CPU technology out-paces the current architectures.

The Radeon M 5650 with 1GB DDR5 128bit was the only exception, fed by a i7-Q 720 mobile CPU and was included for comparison.

GPU Test – The Process

The GPU portion of the benchmark is using a variation of the models seen in SPECviewperf 11 and SPECapc for Maya® 2009, and include X different animated scenes with polygon counts varying from 6 to 66M. Each scene is tested in different display modes: Wireframe / Shaded / Shaded HQ / Shaded with selected geometry outlined. The time needed to complete the fixed frame count in the animated sequence is recorded and stored, so that it can be compared to the reference system.

For reference, this is a real time viewport capture of a single run under SPECapc for Maya® 2012 using a Quadro 600. Check more benchmark captures and watch them in full HD through the PCFoo.com Youtube Channel.

 

 

 

In more detail the model scenes are the following – click on images for full size PNGs:

  • test1 – handShadedhand1.masmoothShadedGrab
    • 33M vertices
    • drawArray





  • test2 – handShaded HQhand1.masmoothShadedHQGrab
    • 66M vertices
    • draw arrays
    • vertex and fragment shaders




  • test3 – hand wireframehand1.mawireframeGrab
    • 22M vertices
    • vertex arrays






  • test4 – squid shadedSquid.masmoothShadedGrab
    • 38M vertices
    • vertex arrays






  • test5 – squid shaded HQSquid.matexturedHQGrab
    • 2M vertices
    • vertex arrays
    • vertex and fragment shaders




  • test6 – toy store shadedToyStoreSPEC.mbtexturedGrab
    • 6M vertices
    • vertex array
    • clipping





  • test7 – toy store wireframeToyStoreSPEC.mbwireframeGrab
    • 9M vertices
    • immediate mode/vertex array
    • clipping




  • test8 – wolf fur selectwerewolfDuet.matexturedSelectedGrab
    • 6.6M vertices
    • immediate mode/vertex array





  • test9 – wolf shaded werewolfDuet.masmoothShadedHQGrab
    • 12M vertices
    • vertex arrays
    • texture





  • test10 – wolf shaded HQwerewolfDuet.matexturedHQGrab
    • 32M vertices
    • vertex arrays
    • vertex and fragment shaders

 

2 thoughts on “SPECapc – Maya® 2012 GPU Scores

  1. Hi Dimitris, Thanks for sharing these tests. Do you have any idea or have tested shave and Lbrush (www.joealter.com) with K2000. I am planning to buy Lenovo W530 for my mobile purpose with 32GB ram and K2000 . It would be great if you can advice me on it. The major purpose will be Maya Lbrush and Shave, view port related and really less rendering, so bacically related to Open GL and some Open CL related stuff. Keep on the great work of sharing knowledge 🙂 . Regards, Sudhir

  2. Yeah, I was working on the same principle, but it just kept crashing (it has some methods which are not thread-safe, and evidently I was using those). How is the GPU process different from multi threading in terms of the code workflow? I would assume that it is the same, except for the API calls to CUDA.

Leave a Reply

Your email address will not be published. Required fields are marked *