AMD FirePro W9100 vs NVIDIA Quadro K6000

3ds Max's 2011 test is a bit long in the tooth (as is this version of the application), but later versions of the program don't support the entire benchmark. We've included two sets of results -- first the Large Model Composite and GPU composite from the SPEC professional benchmark, followed by the individual breakouts for four specific tests. First, here's the high level results:
Note that while the Quadro K6000 wins this test overall in both categories, AMD's W9000's GPU Composite score is much faster now (29.4 fps) than it was in 2012 (23.6 fps). Interestingly, the FirePro W9100 is only slightly faster than the W9000, suggesting that the older card is a sort of sweet spot for the test. The K6000 still takes the test overall, but at least the W9000 has pulled even with the Quadro 6000 that used to sweep it.


In the breakout, workload specific tests, we see more granularity. The FirePro W9000 and W9100 are tied in every test, save for Large Model GPU, where the W9100 is 22% faster. The Quadro K6000 remains the fastest overall card in each test, with the Quadro 6000 bringing up the rear.

One thing to note about 3ds Max 2011 is that we also wanted to test some of the custom plugins people have developed for ray tracing or GPU rendering, but were unable to find many that supported CUDA and OpenCL evenly. Obviously i-ray and Octane Render are both Nvidia-only (the former is an Nvidia creation), but even V-ray, while technically supporting OpenCL, didn't actually work with a modern FirePro card.

SolidWorks 2013:

SolidWorks is a solid modeling CAD program that's actually manufactured by Dassault, not the all-encompassing Autodesk. SPEC has its own SolidWorks benchmark test that puts the program through a comprehensive set of scenarios. This is a new test for us -- we didn't include SolidWorks figures back in 2012.

Here, the W9000 and W9100 stumble -- and oddly, the W9000 is faster than the newer card in every category. This benchmark shows the importance of per-GPU driver optimization and ISV collaboration -- without belaboring the point, there's simply no intrinsic reason why an R9 290X-class GPU would ever lose any benchmark to a GTX 480-class Fermi -- yet the Quadro 6000 is faster in every single test.

This is the one benchmark where Nvidia takes AMD to the cleaners, even accounting for the significant price differential between the two cards. The K6000 may be 50% more expensive than the W9100, but it's also 30-50% faster.